规则定义中的AST和运算符优先级 [英] AST and operator precedence in rule definition
问题描述
您好 [¹]
我有一个简单的解析器
它打算解析条件表达式(关系算术运算及其逻辑组合)。
那么它会成功解析A> 5,但是会停止并忽略输入的其余部分,这与我的内容一致。
如何更改 expr _
规则,使其解析整个输入?
#include< cstdint>
#include< boost / spirit / include / qi.hpp>
#include< boost / spirit / include / phoenix.hpp>
#include< boost / spirit / include / phoenix_operator.hpp>
#include< boost / variant / recursive_wrapper.hpp>
命名空间qi = boost :: spirit :: qi;
namespace phx = boost :: phoenix;
///终端
枚举metric_t:std :: uint8_t {A = 0u,B};
const std :: string metric_names [] = {A,B};
struct metrics_parser:boost :: spirit :: qi :: symbols< char,metric_t>
{
metrics_parser()
{
this-> add
(metric_names [A],A)
(metric_names [B],B)
;
}
};
///运算符
struct op_or {};
struct op_and {};
struct op_xor {};
struct op_not {};
struct op_eq {};
struct op_lt {};
struct op_let {};
struct op_gt {};
struct op_get {};
template< typename tag> struct unop;
template< typename tag> struct binop;
///表达式
typedef boost :: variant<
int,
double,
metric_t,
boost :: recursive_wrapper< unop< op_not> >,
boost :: recursive_wrapper< binop< op_and> >,
boost :: recursive_wrapper< binop< op_or> >,
boost :: recursive_wrapper< binop< op_xor> >,
boost :: recursive_wrapper< binop< op_eq> >,
boost :: recursive_wrapper< binop< op_lt> >,
boost :: recursive_wrapper< binop< op_gt> >
> expr;
template< typename tag>
struct binop
{
explicit binop(const expr& l,const expr& r):oper1(l),oper2(r){}
expr oper1,oper2;
};
template< typename tag>
struct unop
{
显式unop(const expr& o):oper1(o){}
expr oper1;
};
struct printer:boost :: static_visitor< void>
{
printer(std :: ostream& os):_os(os){}
std :: ostream& _os;
void operator()(const binop< op_and>& b)const {print(and,b.oper1,b.oper2); }
void operator()(const binop< op_or>& b)const {print(or,b.oper1,b.oper2); }
void operator()(const binop< op_xor>& b)const {print(xor,b.oper1,b.oper2); }
void operator()(const binop< op_eq&& b)const {print(=,b.oper1,b.oper2); }
void operator()(const binop< op_lt>& b)const {print(<,b.oper1,b.oper2); }
void operator()(const binop< op_gt>& b)const {print(>,b.oper1,b.oper2); }
void print(const std :: string& op,const expr& l,const expr& r)const
{
_os< (;
boost :: apply_visitor(* this,l);
_os<< op;
boost :: apply_visitor(* this,r);
_os <<);
}
void operator()(const unop< op_not>& u)const
{
_os& (;
_os<<!;
boost :: apply_visitor(* this,u.oper1);
_os<
}
void operator()(metric_t m)const
{
_os< metric_names [m];
}
template< typename other_t>
void operator()(other_t i)const
{
_os<<一世;
}
};
std :: ostream& operator<<<(std :: ostream& os,const expr& e)
{boost :: apply_visitor(printer(os),e); return os; }
std :: ostream& operator<<(std :: ostream& os,metric_t m)
{os< metric_names [m]; return os; }
template< typename It,typename Skipper = qi :: space_type>
struct parser:qi :: grammar< It,expr(),Skipper>
{
parser():parser :: base_type(expr_)
{
使用命名空间qi;
using namespace phx;
使用local_names :: _ a;
number_r_%= int_ |双_;
metric_r_%= metric_p_;
eq_r_ =
(metric_r_>>=>> number_r_)
[_val = phx :: construct< binop< op_eq> (_1,_2)] |
(metric_r_>>!=>> number_r_)
[_val = phx :: construct& unop< op_not> >(phx :: construct< binop< op_eq>(_ 1,_2))]
;
ineq_r_ =
(metric_r_>>>>>>< number_r_)
[_val = phx :: construct< binop< op_gt> (_1,_2)] |
(metric_r_>><>> number_r_)
[_val = phx :: construct< binop< op_lt> (_1,_2)] |
(metric_r_>>> =>> number_r_)
[_val = phx :: construct< binop< op_or> >(
phx :: construct< binop< op_gt>>(_ 1,_2),
phx :: construct< binop< op_eq>(_ 1,_2))
] |
(metric_r_>>< =>> number_r_)
[_val = phx :: construct< binop< op_or> >(
phx :: construct< binop< op_lt>>(_ 1,_2),
phx :: construct< binop< op_eq>(_ 1,_2))
]
;
ineq_2_r_ =
(number_r_>><>> metric_r_>><>> phx :: construct& binop< op_and> >(
phx :: construct< binop< op_gt>>(_ 2,_1),
phx :: construct< binop< op_lt>(_ 2,_3))
] |
(number_r_>>< =>>> metric_r_>><>> number_r_)
[_val = phx :: construct< binop< op_and> >(
phx :: construct< binop< op_or>>(
phx :: construct< binop< op_gt>>(_ 2,_1),
phx :: construct< binop< ; op_eq>(_ 2,_1)
),
phx :: construct< binop< op_lt>(_ 2,_3))
]
(number_r_>><>>" metric_r_>>< =>>>>" number_r_)
[_val = phx :: construct< binop< op_and> >(
phx :: construct< binop< op_gt>>(_ 2,_1),
phx :: construct< binop< op_or>>(
phx :: construct< ; op_eq>(_ 2,_3),
phx :: construct< binop< op_lt>(_ 2,_3))
)
]
(number_r_>>< =>> metric_r_>>< =>> number_r_)
[_val = phx :: construct& binop< op_and> >(
phx :: construct< binop< op_or>>(
phx :: construct< binop< op_eq>>(_ 2,_1),
phx :: construct< ; op_gt>>(_ 2,_1)
),
phx :: construct< binop< op_or>(
phx :: construct< binop< op_eq> _3),
phx :: construct< binop< op_lt>>(_ 2,_3)
)
)
]
expr_ =
eq_r_ [_val = _1] |
ineq_r_ [_val = _1] |
ineq_2_r_ [_val = _1] |
(not>> expr_)[_val = phx :: construct< unop< op_not> >(_ 1)] |
(expr_>>和>> expr_)[_val = phx :: construct< binop< op_and> (_1,_2)] |
(expr_>>或>> expr_)[_val = phx :: construct< binop< op_or> (_1,_2)] |
(expr_>>xor>> expr_)[_val = phx :: construct< binop< op_xor> >(_ 1,_2)];
metric_r_.name(metric r);
eq_r_.name(eq_r_);
ineq_r_.name(ineq_r_);
ineq_2_r_.name(ineq_2_r_);
expr_.name(expr_);
debug(metric_r_);
debug(eq_r_);
debug(ineq_r_);
debug(ineq_2_r_);
debug(expr_);
}
private:
metrics_parser metric_p_;
qi :: rule< It,expr(),Skipper> number_r_;
qi :: rule< It,expr(),Skipper> metric_r_;
qi :: rule< It,expr(),Skipper> eq_r_;
qi :: rule< It,expr(),Skipper> ineq_r_;
qi :: rule< It,expr(),Skipper> ineq_2_r_;
qi :: rule< It,expr(),Skipper> expr_;
};
int main()
{
std :: list< std :: string> lstr;
lstr.emplace_back(A> 5且B <4 xor A> 3.4或2
for(auto i = std :: begin(lstr); i!= std :: end(lstr); ++ i)
{
auto& input = * i;
auto f(std :: begin(input)),l(std :: end(input));
parser< decltype(f)> p;
try
{
expr result;
bool ok = qi :: phrase_parse(f,l,p,qi :: space,result);
if(!ok)
std :: cerr<< invalid input\\\
;
else
std :: cout<< result:<结果< \\\
;
} catch(const qi :: expectation_failure< decltype(f)>& e)
{
std :: cerr< expectation_failure at'< std :: string(e.first,e.last)<< '\\\
;
}
if(f!= l)std :: cerr<< unparsed:'< std :: string(f,l)<< '\\\
;
}
return 0;
}
MM
[¹] 问题从 [精神]用户列表
$
relop_expr = eq_r_ | ineq_r_ | ineq_2_r_;
expr_ =
(not>> expr_)[_val = phx :: construct< unop< op_not> >(_ 1)] |
(relop_expr>>和>> expr_)[_val = phx :: construct< binop< op_and> (_1,_2)] |
(relop_expr>>或>> expr_)[_val = phx :: construct< binop< op_or> (_1,_2)] |
(relop_expr>>xor>> expr_)[_val = phx :: construct< binop< op_xor> (_1,_2)] |
(relop_expr)[_val = _1]
;
BOOST_SPIRIT_DEBUG_NODES((metric_r _)(eq_r _)(ineq_r _)(ineq_2_r _)(relop_expr)(expr_))
注意:
- 分支的顺序
- 一个额外的级别(
relop_expr
)以诱导优先
( 3.4
尚未解析, 2 也没有解析)。此外,它是令人难以置信的低效率(可以做左因子分解)。修正这些:
number_r_ = real_parser< double,strict_real_policies< double>> int_;
relop_expr = eq_r_ | ineq_2_r_ | ineq_r_;
expr_ =
(not>> expr_)[_val = construct< unop< op_not> (_1)] |
relop_expr [_a = _1]>> (
(和>> expr_ [_val = bin_< op_and>()])|
(or>> expr_ [_val = bin_< op_or> )|
(xor>> expr_ [_val = bin_< op_xor>()])|
(eps [_val = _a])
)
正如你所看到的,我真的不能满足那些复杂的语义动作。这个的主要原因是BUGS。使代码可读,丢失一半的错误。因此,只需两个简单的帮助,我们可以减少冗长:
模板< typename标签>使用bin_ = decltype(phx :: construct< binop< Tag>>(qi :: _a,qi :: _ 1));
template< typename T1,typename T2>使用tern_ = decltype(phx :: construct< binop< op_and>>(phx :: construct< binop< T1>>(qi :: _ a,qi :: _1),phx :: construct< binop& ;(qi :: 1,qi :: 2)));
正如你所看到的,我不会努力写出特质等等。对
4非常干净的行:
ineq_2_r_ = number_r_ [_a = _1]> (
(<>> metric_r_>><>> number_r_)[_val = tern_< op_lt,op_lt>()] |
>>" metric_r_>>< =>> number_r_)[_val = tern_< op_lt,op_lte>()] |
(< => metric_r_> ;>>>>>>>>>< = >>>< number_r_)[_val = tern_< op_lte,op_lte>()] |
//看,这很容易,我们甚至可以在奖金中 - 与书写锅炉板:)
(>>> metric_r_>>>>>>>>< number_r_)[_val = tern_< op_gt,op_gt> |
(>>> metric_r_>>> =>> number_r_)[_val = tern_< op_gt,op_gte>()] |
(> =>> metric_r_>>>>> number_r_)[_val = tern_< op_gte,op_gt>()] |
(> =>> metric_r_>>> =>> number_r_)[_val = tern_< op_gte,op_gte>()]
);
哦,我刚刚记住:我已经定义了 op_gte
和 op_lte
运算符,因为没有它们导致语义动作的二次增长。我的经验法则是:
- 规则1 ,避免语义操作
- 推论#1 :使您的AST直接反映语法。
在这种情况下,你使用解析来混合AST转换。如果你想将AST转换为'expand'lte(a,b)< - (lt(a,b)|| eq(a,b)),你可以在解析。 更新 查看演示的其他答案
$ b $总而言之,我在一个工作计划中附上了建议。它实现更多的功能,并在73行更短(28%)。即使有更多的测试用例:
'A> 5':结果:(A> 5)
/ pre>
'A< 5':结果:(A <5)
'A> = 5':结果:(A> = 5)
'A <= 5':result: = 5)
'A = 5':结果:(A = 5)
'A!= 5':结果: (A> 3.4)或2 5)和((B <4) ; 3)))))
'A> 5和B <4 xor A!= 3.4或7.9e10> = B> = -42':结果:((A> 5) B <4)xor(A = 3.4)或((7.9e + 10 = B)和(B> = -42))))
好吧,我会显示它在Coliru上,但它似乎下来了。希望你喜欢这个。
完整示例
//#define BOOST_SPIRIT_DEBUG
#include< boost / spirit / include / qi.hpp>
#include< boost / spirit / include / phoenix.hpp>
#include< boost / spirit / include / phoenix_operator.hpp>
#include< boost / variant / recursive_wrapper.hpp>
#include< cstdint>
命名空间qi = boost :: spirit :: qi;
namespace phx = boost :: phoenix;
///终端
枚举metric_t:std :: uint8_t {A = 0u,B};
const std :: string metric_names [] = {A,B};
struct metrics_parser:boost :: spirit :: qi :: symbols< char,metric_t> {
metrics_parser(){
this-> add(metric_names [A],A)
(metric_names [B],B);
}
};
/// Operators
template< typename tag> struct unop;
template< typename tag> struct binop;
///表达式
typedef boost :: variant<
int,
double,
metric_t,
boost :: recursive_wrapper< unop< struct op_not> >,
boost :: recursive_wrapper< binop< struct op_and> >,
boost :: recursive_wrapper< binop< struct op_or> >,
boost :: recursive_wrapper< binop< struct op_xor> >,
boost :: recursive_wrapper< binop< struct op_eq> >,
boost :: recursive_wrapper< binop< struct op_lt> >,
boost :: recursive_wrapper< Binop< struct op_gt> >,
boost :: recursive_wrapper< binop< struct op_lte> >,
boost :: recursive_wrapper< binop< struct op_gte> >
> expr;
template< typename tag>
struct binop {
explicit binop(const expr& l,const expr& r):oper1(1),oper2(r){}
expr oper1,oper2;
};
template< typename tag>
struct unop {
explicit unop(const expr& o):oper1(o){}
expr oper1;
};
std :: ostream& operator<<<(std :: ostream& os,metric_t m)
{return os< metric_names [m]; }
结构打印机:boost :: static_visitor< void>
{
printer(std :: ostream& os):_os(os){}
std :: ostream& _os;
void operator()(const binop< op_and>& b)const {print(and,b.oper1,b.oper2); }
void operator()(const binop< op_or>& b)const {print(or,b.oper1,b.oper2); }
void operator()(const binop< op_xor>& b)const {print(xor,b.oper1,b.oper2); }
void operator()(const binop< op_eq&& b)const {print(=,b.oper1,b.oper2); }
void operator()(const binop< op_lt>& b)const {print(<,b.oper1,b.oper2); }
void operator()(const binop< op_gt>& b)const {print(>,b.oper1,b.oper2); }
void operator()(const binop< op_lte>& b)const {print(< =,b.oper1,b.oper2); }
void operator()(const binop< op_gte>& b)const {print(> =,b.oper1,b.oper2); }
void print(const std :: string& op,const expr& l,const expr& r)const {
_os& (;
boost :: apply_visitor(* this,l); _os<< op; boost :: apply_visitor(* this,r);
_os<
}
void operator()(const unop< op_not>& u)const {
_os< !; boost :: apply_visitor(* this,u.oper1);
}
template< typename other_t> void operator()(other_t i)const {
_os<<一世;
}
};
std :: ostream& operator<<<(std :: ostream& os,const expr& e)
{boost :: apply_visitor(printer(os),e); return os; }
template< typename It,typename Skipper = qi :: space_type>
struct parser:qi :: grammar< It,expr(),Skipper,qi :: locals< expr> >
{
template< typename Tag>使用bin_ = decltype(phx :: construct< binop< Tag>>(qi :: _a,qi :: _ 1));
template< typename T1,typename T2>使用tern_ = decltype(phx :: construct< binop< op_and>>(phx :: construct< binop< T1>>(qi :: _ a,qi :: _1),phx :: construct< binop& ;(qi :: 1,qi :: 2)));
parser():parser :: base_type(expr_)
{
使用命名空间qi;
using namespace phx;
number_r_ = real_parser< double,strict_real_policies< double>>()| int_;
metric_r_ = metric_p_;
eq_r_ = metric_r_ [_a = _1]>> (
(=>> number_r_)[_val = bin_< op_eq>()] |
(!=>> number_r_)[_val = construct< unop< op_not> >(bin_< op_eq>())]
);
ineq_2_r_ = number_r_ [_a = _1]>> (
(<>> metric_r_>><>> number_r_)[_val = tern_< op_lt,op_lt>()] |
>>" metric_r_>>< =>> number_r_)[_val = tern_< op_lt,op_lte>()] |
(< => metric_r_> ;>>>>>>>>>< = >>>>< number_r_)[_val = tern_< op_lte,op_lte>()] |
(>>> metric_r_>>& tern_< op_gt,op_gt>()] |
(>>>>>>>>>>< number_r_)[_val = tern_< op_gt,op_gte> ] |
(> =>> metric_r_>>>>>< number_r_)[_val = tern_< op_gte,op_gt>()] |
> =>> metric_r_>> =>>>< number_r_)[_val = tern_< op_gte,op_gte>()]
);
ineq_r_ = metric_r_ [_a = _1]>> (
(>>> number_r_)[_val = bin_< op_gt>()] |
(<> number_r_)[_val = bin_< op_lt> ;()] |
(> =>> number_r_)[_val = bin_< op_gte>()] |
(< => number_r_) = bin_< op_lte>()]
);
relop_expr = eq_r_ | ineq_2_r_ | ineq_r_;
expr_ =
(not>> expr_)[_val = construct< unop< op_not> (_1)] |
relop_expr [_a = _1]>>> (
(和>> expr_ [_val = bin_< op_and>()])|
(or>> expr_ [_val = bin_< op_or> )|
(xor> expr_ [_val = bin_< op_xor>()])|
(eps [_val = _a])
);
BOOST_SPIRIT_DEBUG_NODES((metric_r _)(eq_r _)(ineq_r _)(ineq_2_r _)(relop_expr)(expr_))
}
private:
qi :: rule< It ,expr(),Skipper,qi :: locals< expr> > eq_r_,ineq_r_,ineq_2_r_,relop_expr,expr_;
qi :: rule< It,expr(),Skipper> number_r_,metric_r_;
metrics_parser metric_p_;
};
int main()
{
for(std :: string const& input:{
A> 5,
A< ; 5,
A> = 5,
A <= 5,
A = 5,
A!= 5 b $ bA> 5且B <4或A> 3.4或2 A> 5且B <4或A 0 = 3.4或7.9e 10> = B> -42
})
{
auto f(std :: begin(input)),l(std :: end(input));
parser< decltype(f)> p;
try
{
std :: cout< '<输入<< ':\t;
expr result;
bool ok = qi :: phrase_parse(f,l,p,qi :: space,result);
if(!ok)std :: cout<< invalid input\\\
;
else std :: cout<< result:<结果< \\\
;
} catch(const qi :: expectation_failure< decltype(f)>& e)
{
std :: cout< expectation_failure at'< std :: string(e.first,e.last)<< '\\\
;
}
if(f!= l)std :: cout<< unparsed:'< std :: string(f,l)<< '\\\
;
}
}
Hello [¹]
I have a simple parser (see below).
It intends to parse conditional expressions (relational arithmetic operations and logic combinations thereof).
In the example given there, it parses successfully A>5 but then stops and ignores the rest of the input, and this is consistent with my impl.
How do I change the
expr_
rule to make it parse the entire input?#include <cstdint> #include <boost/spirit/include/qi.hpp> #include <boost/spirit/include/phoenix.hpp> #include <boost/spirit/include/phoenix_operator.hpp> #include <boost/variant/recursive_wrapper.hpp> namespace qi = boost::spirit::qi; namespace phx = boost::phoenix; /// Terminals enum metric_t : std::uint8_t { A=0u, B }; const std::string metric_names[] = { "A", "B" }; struct metrics_parser : boost::spirit::qi::symbols<char, metric_t> { metrics_parser() { this->add ( metric_names[A], A ) ( metric_names[B], B ) ; } }; /// Operators struct op_or {}; struct op_and {}; struct op_xor {}; struct op_not {}; struct op_eq {}; struct op_lt {}; struct op_let {}; struct op_gt {}; struct op_get {}; template <typename tag> struct unop; template <typename tag> struct binop; /// Expression typedef boost::variant< int, double, metric_t, boost::recursive_wrapper< unop<op_not> >, boost::recursive_wrapper< binop<op_and> >, boost::recursive_wrapper< binop<op_or> >, boost::recursive_wrapper< binop<op_xor> >, boost::recursive_wrapper< binop<op_eq> >, boost::recursive_wrapper< binop<op_lt> >, boost::recursive_wrapper< binop<op_gt> > > expr; template <typename tag> struct binop { explicit binop(const expr& l, const expr& r) : oper1(l), oper2(r) { } expr oper1, oper2; }; template <typename tag> struct unop { explicit unop(const expr& o) : oper1(o) { } expr oper1; }; struct printer : boost::static_visitor<void> { printer(std::ostream& os) : _os(os) {} std::ostream& _os; void operator()(const binop<op_and>& b) const { print(" and ", b.oper1, b.oper2); } void operator()(const binop<op_or >& b) const { print(" or ", b.oper1, b.oper2); } void operator()(const binop<op_xor>& b) const { print(" xor ", b.oper1, b.oper2); } void operator()(const binop<op_eq>& b) const { print(" = ", b.oper1, b.oper2); } void operator()(const binop<op_lt>& b) const { print(" < ", b.oper1, b.oper2); } void operator()(const binop<op_gt>& b) const { print(" > ", b.oper1, b.oper2); } void print(const std::string& op, const expr& l, const expr& r) const { _os << "("; boost::apply_visitor(*this, l); _os << op; boost::apply_visitor(*this, r); _os << ")"; } void operator()(const unop<op_not>& u) const { _os << "("; _os << "!"; boost::apply_visitor(*this, u.oper1); _os << ")"; } void operator()(metric_t m) const { _os << metric_names[m]; } template <typename other_t> void operator()(other_t i) const { _os << i; } }; std::ostream& operator<<(std::ostream& os, const expr& e) { boost::apply_visitor(printer(os), e); return os; } std::ostream& operator<<(std::ostream& os, metric_t m) { os<< metric_names[m]; return os; } template <typename It, typename Skipper = qi::space_type> struct parser : qi::grammar<It, expr(), Skipper> { parser() : parser::base_type(expr_) { using namespace qi; using namespace phx; using local_names::_a; number_r_ %= int_ | double_; metric_r_ %= metric_p_; eq_r_ = (metric_r_ >> "=" >> number_r_) [ _val = phx::construct< binop<op_eq> >(_1,_2) ] | (metric_r_ >> "!=" >> number_r_) [ _val = phx::construct< unop<op_not> >( phx::construct< binop<op_eq> >(_1,_2) ) ] ; ineq_r_ = (metric_r_ >> ">" >> number_r_) [ _val = phx::construct< binop<op_gt> >(_1,_2) ] | (metric_r_ >> "<" >> number_r_) [ _val = phx::construct< binop<op_lt> >(_1,_2) ] | (metric_r_ >> ">=" >> number_r_) [ _val = phx::construct< binop<op_or> >( phx::construct< binop<op_gt> >(_1,_2), phx::construct< binop<op_eq> >(_1,_2) ) ] | (metric_r_ >> "<=" >> number_r_) [ _val = phx::construct< binop<op_or> >( phx::construct< binop<op_lt> >(_1,_2), phx::construct< binop<op_eq> >(_1,_2) ) ] ; ineq_2_r_ = (number_r_ >> "<" >> metric_r_ >> "<" >> number_r_) [ _val = phx::construct< binop<op_and> >( phx::construct< binop<op_gt> >(_2,_1), phx::construct< binop<op_lt> >(_2,_3) ) ] | (number_r_ >> "<=" >> metric_r_ >> "<" >> number_r_) [ _val = phx::construct< binop<op_and> >( phx::construct< binop<op_or> >( phx::construct< binop<op_gt> >(_2,_1), phx::construct< binop<op_eq> >(_2,_1) ), phx::construct< binop<op_lt> >(_2,_3) ) ] | (number_r_ >> "<" >> metric_r_ >> "<=" >> number_r_) [ _val = phx::construct< binop<op_and> >( phx::construct< binop<op_gt> >(_2,_1), phx::construct< binop<op_or> >( phx::construct< binop<op_eq> >(_2,_3), phx::construct< binop<op_lt> >(_2,_3) ) ) ] | (number_r_ >> "<=" >> metric_r_ >> "<=" >> number_r_) [ _val = phx::construct< binop<op_and> >( phx::construct< binop<op_or> >( phx::construct< binop<op_eq> >(_2,_1), phx::construct< binop<op_gt> >(_2,_1) ), phx::construct< binop<op_or> >( phx::construct< binop<op_eq> >(_2,_3), phx::construct< binop<op_lt> >(_2,_3) ) ) ] ; expr_ = eq_r_ [ _val = _1 ] | ineq_r_ [ _val = _1 ] | ineq_2_r_ [ _val = _1 ] | ("not" >> expr_) [ _val = phx::construct< unop<op_not> >(_1) ] | (expr_ >> "and" >> expr_) [ _val = phx::construct< binop<op_and> >(_1,_2) ] | (expr_ >> "or" >> expr_) [ _val = phx::construct< binop<op_or> >(_1,_2) ] | (expr_ >> "xor" >> expr_) [ _val = phx::construct< binop<op_xor> >(_1,_2) ]; metric_r_.name("metric r"); eq_r_.name("eq_r_"); ineq_r_.name("ineq_r_"); ineq_2_r_.name("ineq_2_r_"); expr_.name("expr_"); debug(metric_r_); debug(eq_r_); debug(ineq_r_); debug(ineq_2_r_); debug(expr_); } private: metrics_parser metric_p_; qi::rule<It, expr(), Skipper> number_r_; qi::rule<It, expr(), Skipper> metric_r_; qi::rule<It, expr(), Skipper> eq_r_; qi::rule<It, expr(), Skipper> ineq_r_; qi::rule<It, expr(), Skipper> ineq_2_r_; qi::rule<It, expr(), Skipper> expr_; }; int main() { std::list<std::string> lstr; lstr.emplace_back("A>5 and B<4 xor A>3.4 or 2<A<3"); for (auto i=std::begin(lstr); i!=std::end(lstr); ++i) { auto& input = *i; auto f(std::begin(input)), l(std::end(input)); parser<decltype(f)> p; try { expr result; bool ok = qi::phrase_parse(f,l,p,qi::space,result); if (!ok) std::cerr << "invalid input\n"; else std::cout << "result: " << result << "\n"; } catch (const qi::expectation_failure<decltype(f)>& e) { std::cerr << "expectation_failure at '" << std::string(e.first, e.last) << "'\n"; } if (f!=l) std::cerr << "unparsed: '" << std::string(f,l) << "'\n"; } return 0; }
Thanks, MM
[¹] question teleported from the [spirit-general] user list
解决方案Sticking with simple:
relop_expr = eq_r_ | ineq_r_ | ineq_2_r_; expr_ = ("not" >> expr_) [ _val = phx::construct< unop<op_not> >(_1) ] | (relop_expr >> "and" >> expr_) [ _val = phx::construct< binop<op_and> >(_1,_2) ] | (relop_expr >> "or" >> expr_) [ _val = phx::construct< binop<op_or> >(_1,_2) ] | (relop_expr >> "xor" >> expr_) [ _val = phx::construct< binop<op_xor> >(_1,_2) ] | (relop_expr ) [ _val = _1 ] ; BOOST_SPIRIT_DEBUG_NODES((metric_r_)(eq_r_)(ineq_r_)(ineq_2_r_)(relop_expr)(expr_))
Note:
- the ordering of branches
- the use of an extra "level" (
relop_expr
) to induce precedence
There's still work to do (3.4
did not parse yet, and neither did 2<A<3
). Also, it's excruciatingly inefficient (could do with left factorization). Fixing those:
number_r_ = real_parser<double,strict_real_policies<double>>() | int_;
relop_expr = eq_r_ | ineq_2_r_ | ineq_r_;
expr_ =
("not" >> expr_) [ _val = construct<unop<op_not>> (_1) ] |
relop_expr [_a = _1] >> (
("and" >> expr_ [ _val = bin_<op_and>() ]) |
("or" >> expr_ [ _val = bin_<op_or >() ]) |
("xor" >> expr_ [ _val = bin_<op_xor>() ]) |
(eps [ _val = _a ])
)
;
As you can see, I can't really stand those complicated semantic actions. The chief reason for this is BUGS. Make the code readable, lose half the bugs. So, with just a two simple helpers we can reduce the verbosity:
template <typename Tag> using bin_ = decltype(phx::construct<binop<Tag>>(qi::_a, qi::_1));
template <typename T1, typename T2> using tern_ = decltype(phx::construct<binop<op_and>>(phx::construct<binop<T1>>(qi::_a, qi::_1), phx::construct<binop<T2>>(qi::_1, qi::_2)));
As you can see, I don't make a great effort to write traits etc. Just a quick decltype on whatever you'd write anyways, and, bam
down from 35 crufty lines to 4 very clean lines:
ineq_2_r_ = number_r_ [ _a = _1 ] >> (
("<" >> metric_r_ >> "<" >> number_r_) [_val = tern_<op_lt , op_lt>() ] |
("<" >> metric_r_ >> "<=" >> number_r_) [_val = tern_<op_lt , op_lte>() ] |
("<=" >> metric_r_ >> "<" >> number_r_) [_val = tern_<op_lte, op_lt>() ] |
("<=" >> metric_r_ >> "<=" >> number_r_) [_val = tern_<op_lte, op_lte>() ] |
// see, that's so easy, we can even trow in the bonus - I bet you were just fed up with writing boiler plate :)
(">" >> metric_r_ >> ">" >> number_r_) [_val = tern_<op_gt , op_gt>() ] |
(">" >> metric_r_ >> ">=" >> number_r_) [_val = tern_<op_gt , op_gte>() ] |
(">=" >> metric_r_ >> ">" >> number_r_) [_val = tern_<op_gte, op_gt>() ] |
(">=" >> metric_r_ >> ">=" >> number_r_) [_val = tern_<op_gte, op_gte>() ]
);
Oh, I just remembered: I have defined the op_gte
and op_lte
operators, since not having them was causing quadratic growth of your semantic actions. My fast rule of thumb is:
- Rule #1: keep rules simple, avoid semantic actions
- Corollary #1: make your AST directly reflect the grammar.
In this case, you were conflating AST transformation with parsing. If you want to transform the AST to 'expand' lte (a,b) <- (lt(a,b) || eq(a,b)), you can trivially do that after parsing. Update see the other answer for a demo
All in all, I have attached the suggestions in a working program. It implements many more features, and comes in 73 lines shorter (28%). That's even with more test cases:
'A > 5': result: (A > 5)
'A < 5': result: (A < 5)
'A >= 5': result: (A >= 5)
'A <= 5': result: (A <= 5)
'A = 5': result: (A = 5)
'A != 5': result: !(A = 5)
'A>5 and B<4 xor A>3.4 or 2<A<3': result: ((A > 5) and ((B < 4) xor ((A > 3.4) or ((2 < A) and (A < 3)))))
'A>5 and B<4 xor A!=3.4 or 7.9e10 >= B >= -42': result: ((A > 5) and ((B < 4) xor (!(A = 3.4) or ((7.9e+10 >= B) and (B >= -42)))))
Well, I'd have shown it live on Coliru, but it seems down at the moment. Hope you like this.
Full sample
//#define BOOST_SPIRIT_DEBUG
#include <boost/spirit/include/qi.hpp>
#include <boost/spirit/include/phoenix.hpp>
#include <boost/spirit/include/phoenix_operator.hpp>
#include <boost/variant/recursive_wrapper.hpp>
#include <cstdint>
namespace qi = boost::spirit::qi;
namespace phx = boost::phoenix;
/// Terminals
enum metric_t : std::uint8_t { A=0u, B };
const std::string metric_names[] = { "A", "B" };
struct metrics_parser : boost::spirit::qi::symbols<char, metric_t> {
metrics_parser() {
this->add(metric_names[A], A)
(metric_names[B], B);
}
};
/// Operators
template <typename tag> struct unop;
template <typename tag> struct binop;
/// Expression
typedef boost::variant<
int,
double,
metric_t,
boost::recursive_wrapper< unop< struct op_not> >,
boost::recursive_wrapper< binop<struct op_and> >,
boost::recursive_wrapper< binop<struct op_or> >,
boost::recursive_wrapper< binop<struct op_xor> >,
boost::recursive_wrapper< binop<struct op_eq> >,
boost::recursive_wrapper< binop<struct op_lt> >,
boost::recursive_wrapper< binop<struct op_gt> >,
boost::recursive_wrapper< binop<struct op_lte> >,
boost::recursive_wrapper< binop<struct op_gte> >
> expr;
template <typename tag>
struct binop {
explicit binop(const expr& l, const expr& r) : oper1(l), oper2(r) { }
expr oper1, oper2;
};
template <typename tag>
struct unop {
explicit unop(const expr& o) : oper1(o) { }
expr oper1;
};
std::ostream& operator<<(std::ostream& os, metric_t m)
{ return os << metric_names[m]; }
struct printer : boost::static_visitor<void>
{
printer(std::ostream& os) : _os(os) {}
std::ostream& _os;
void operator()(const binop<op_and>& b) const { print(" and ", b.oper1, b.oper2); }
void operator()(const binop<op_or >& b) const { print(" or ", b.oper1, b.oper2); }
void operator()(const binop<op_xor>& b) const { print(" xor ", b.oper1, b.oper2); }
void operator()(const binop<op_eq >& b) const { print(" = ", b.oper1, b.oper2); }
void operator()(const binop<op_lt >& b) const { print(" < ", b.oper1, b.oper2); }
void operator()(const binop<op_gt >& b) const { print(" > ", b.oper1, b.oper2); }
void operator()(const binop<op_lte>& b) const { print(" <= ", b.oper1, b.oper2); }
void operator()(const binop<op_gte>& b) const { print(" >= ", b.oper1, b.oper2); }
void print(const std::string& op, const expr& l, const expr& r) const {
_os << "(";
boost::apply_visitor(*this, l); _os << op; boost::apply_visitor(*this, r);
_os << ")";
}
void operator()(const unop<op_not>& u) const {
_os << "!"; boost::apply_visitor(*this, u.oper1);
}
template <typename other_t> void operator()(other_t i) const {
_os << i;
}
};
std::ostream& operator<<(std::ostream& os, const expr& e)
{ boost::apply_visitor(printer(os), e); return os; }
template <typename It, typename Skipper = qi::space_type >
struct parser : qi::grammar<It, expr(), Skipper, qi::locals<expr> >
{
template <typename Tag> using bin_ = decltype(phx::construct<binop<Tag>>(qi::_a, qi::_1));
template <typename T1, typename T2> using tern_ = decltype(phx::construct<binop<op_and>>(phx::construct<binop<T1>>(qi::_a, qi::_1), phx::construct<binop<T2>>(qi::_1, qi::_2)));
parser() : parser::base_type(expr_)
{
using namespace qi;
using namespace phx;
number_r_ = real_parser<double,strict_real_policies<double>>() | int_;
metric_r_ = metric_p_;
eq_r_ = metric_r_ [ _a = _1 ] >> (
("=" >> number_r_) [ _val = bin_<op_eq>() ] |
("!=" >> number_r_) [ _val = construct<unop<op_not>>(bin_<op_eq>()) ]
);
ineq_2_r_ = number_r_ [ _a = _1 ] >> (
("<" >> metric_r_ >> "<" >> number_r_) [_val = tern_<op_lt , op_lt>() ] |
("<" >> metric_r_ >> "<=" >> number_r_) [_val = tern_<op_lt , op_lte>() ] |
("<=" >> metric_r_ >> "<" >> number_r_) [_val = tern_<op_lte, op_lt>() ] |
("<=" >> metric_r_ >> "<=" >> number_r_) [_val = tern_<op_lte, op_lte>() ] |
(">" >> metric_r_ >> ">" >> number_r_) [_val = tern_<op_gt , op_gt>() ] |
(">" >> metric_r_ >> ">=" >> number_r_) [_val = tern_<op_gt , op_gte>() ] |
(">=" >> metric_r_ >> ">" >> number_r_) [_val = tern_<op_gte, op_gt>() ] |
(">=" >> metric_r_ >> ">=" >> number_r_) [_val = tern_<op_gte, op_gte>() ]
);
ineq_r_ = metric_r_ [ _a = _1 ] >> (
(">" >> number_r_) [ _val = bin_<op_gt >() ] |
("<" >> number_r_) [ _val = bin_<op_lt >() ] |
(">=" >> number_r_) [ _val = bin_<op_gte>() ] |
("<=" >> number_r_) [ _val = bin_<op_lte>() ]
);
relop_expr = eq_r_ | ineq_2_r_ | ineq_r_;
expr_ =
("not" >> expr_) [ _val = construct<unop<op_not>> (_1) ] |
relop_expr [_a = _1] >> (
("and" >> expr_ [ _val = bin_<op_and>() ]) |
("or" >> expr_ [ _val = bin_<op_or >() ]) |
("xor" >> expr_ [ _val = bin_<op_xor>() ]) |
(eps [ _val = _a ])
);
BOOST_SPIRIT_DEBUG_NODES((metric_r_)(eq_r_)(ineq_r_)(ineq_2_r_)(relop_expr)(expr_))
}
private:
qi::rule<It, expr(), Skipper, qi::locals<expr> > eq_r_, ineq_r_, ineq_2_r_, relop_expr, expr_;
qi::rule<It, expr(), Skipper> number_r_, metric_r_;
metrics_parser metric_p_;
};
int main()
{
for (std::string const& input : {
"A > 5",
"A < 5",
"A >= 5",
"A <= 5",
"A = 5",
"A != 5",
"A>5 and B<4 xor A>3.4 or 2<A<3",
"A>5 and B<4 xor A!=3.4 or 7.9e10 >= B >= -42"
})
{
auto f(std::begin(input)), l(std::end(input));
parser<decltype(f)> p;
try
{
std::cout << "'" << input << "':\t";
expr result;
bool ok = qi::phrase_parse(f,l,p,qi::space,result);
if (!ok) std::cout << "invalid input\n";
else std::cout << "result: " << result << "\n";
} catch (const qi::expectation_failure<decltype(f)>& e)
{
std::cout << "expectation_failure at '" << std::string(e.first, e.last) << "'\n";
}
if (f!=l) std::cout << "unparsed: '" << std::string(f,l) << "'\n";
}
}
这篇关于规则定义中的AST和运算符优先级的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!