n-ary布尔语法从中缀到前缀的Boost::Spirit转换?

n-ary boolean grammar conversion from infix to prefix with Boost::Spirit?

本文关键字：Boost Spirit 转换 前缀 语法 布尔 中缀 n-ary      更新时间：2023-10-16

我需要使用 Boost：：Spirit 将如下所示的中缀符号转换为 n 元前缀表示法，但我未能在 https://stackoverflow.com/a/8707598/1816477 等人的答案的基础上构建。

这是我试图解析的内容：

not (xyz='a' or xyz='b' or xyz='c') and abc='s' xor (pqr ='v' and xyz='d')

这种 LISP 风格的格式是我试图作为输出提供的(不介意缩进(：

(xor (and (= pqr 'v') (= xyz 'd'))
(and (= abc 's')
(not (or (= xyz 'a')
(= xyz 'b')
(= xyz 'c')))))

因此，我尝试解析的术语由前缀(not <expression>(和中缀表达式(<expression> and <expression> and ...等(组成，即：赋值，否定和n元和，or，xors等，暗示运算符优先级(或<和><赋值><否定(。>

我失败的是语法正确。输出到一个合适的boost::variant表示我认为我能够完成的解析布尔表达式。我正在考虑这样的输出结构：

struct prefixExpr;
struct infixExpr;
typedef boost::variant<
std::string,    // identifiers, values etc.
boost::recursive_wrapper<prefixExpr>,   // e.g. negation
boost::recursive_wrapper<infixExpr>     // assignment, and, or, xor etc.
> expression;
struct prefixExpr {
std::string op;    // currently only "not"
expression expr;
};
BOOST_FUSION_ADAPT_STRUCT(prefixExpr, op, expr)
struct infixExpr {
std::string op;    // "and", "or", "xor", "="
std::vector<expression> exprs;
};
BOOST_FUSION_ADAPT_STRUCT(infixExpr, op, exprs)

我需要做什么才能解析上面提到的表达式并将它们转换为前缀表示法？

我正在使用boost 1.67.0(撰写本文时的最新版本(和Visual Studio 15.7.3(在撰写本文时也是最新的(。

代码并不完美，但应该易于理解：

#include <boost/variant.hpp>
#include <boost/spirit/home/x3.hpp>
#include <vector>
#include <string>
#include <iostream>

struct id : std::string {};
struct value : std::string {};
struct nary_expr;
using expr = boost::variant<
id, value,
boost::recursive_wrapper<nary_expr>
>;

struct nary_expr
{
std::string op;
std::vector<expr> exprs;
};

namespace x3 = boost::spirit::x3;
auto compose_nary_expr = [](auto& ctx)
{
//auto&& [left, tail] = x3::_attr(ctx);
auto&& left = boost::fusion::at_c<0>(x3::_attr(ctx));
auto&& tail = boost::fusion::at_c<1>(x3::_attr(ctx));
if (tail.size() == 0) {
x3::_val(ctx) = left;
return;
}
// left associativity
auto op = boost::fusion::at_c<0>(tail[0]);
std::vector<expr> exprs = { left, boost::fusion::at_c<1>(tail[0]) };
for (std::size_t i = 1; i < tail.size(); ++i) {
// same priority but different operator
auto&& next_op = boost::fusion::at_c<0>(tail[i]);
if (op != next_op) {
exprs = std::vector<expr>{ nary_expr{ op, std::move(exprs) } };
op = next_op;
}
exprs.push_back(boost::fusion::at_c<1>(tail[i]));
}
x3::_val(ctx) = nary_expr{ op, std::move(exprs) };
};
x3::rule<class prec4_expr_rule, expr> const prec4_expr("prec4_expr");
x3::rule<class prec3_expr_rule, expr> const prec3_expr("prec3_expr");
x3::rule<class prec2_expr_rule, expr> const prec2_expr("prec2_expr");
x3::rule<class prec1_expr_rule, expr> const prec1_expr("prec1_expr");
x3::rule<class prec0_expr_rule, expr> const prec0_expr("prec0_expr");
auto const prec4_expr_def = prec4_expr = (
prec3_expr
>> *(   (x3::string("or") > prec3_expr)
)
)[compose_nary_expr];
auto const prec3_expr_def = prec3_expr = (
prec2_expr
>> *(   (x3::string("xor") > prec2_expr)
)
)[compose_nary_expr];
auto const prec2_expr_def = prec2_expr = (
prec1_expr
>> *(   (x3::string("and") > prec1_expr)
)
)[compose_nary_expr];

auto compose_binary_expr = [](auto& ctx)
{
auto&& rhs = boost::fusion::at_c<0>(x3::_attr(ctx));
auto&& tail = boost::fusion::at_c<1>(x3::_attr(ctx));
if (tail.size() > 0) {
auto&& op = boost::fusion::at_c<0>(tail[0]);
auto&& lhs = boost::fusion::at_c<1>(tail[0]);
x3::_val(ctx) = nary_expr{ op, { rhs, lhs } };
}
else {
x3::_val(ctx) = rhs;
}
};

// should use optional, but something wrong with spirit
auto const prec1_expr_def = prec1_expr = (
prec0_expr >> *(x3::string("=") > prec0_expr)
)[compose_binary_expr];

x3::rule<class not_expr_rule, expr> const not_expr("not_expr");
auto compose_unary_expr = [](auto& ctx)
{
//auto&& [op, expr] = x3::_attr(ctx);
auto&& op = boost::fusion::at_c<0>(x3::_attr(ctx));
auto&& expr = boost::fusion::at_c<1>(x3::_attr(ctx));
x3::_val(ctx) = nary_expr{ op, { expr } };
};
auto const not_expr_def = not_expr = (x3::string("not") > prec0_expr)[compose_unary_expr];
auto const id_term = x3::rule<class id_r, id>{} = x3::lexeme[x3::alpha >> *x3::alnum];
auto const value_term = x3::rule<class value_r, value>{} = x3::lexeme["'" > +~x3::char_(''') >> "'"];
auto const prec0_expr_def =
value_term
| ( '(' > prec4_expr >> ')' )
| not_expr
| id_term
;

BOOST_SPIRIT_DEFINE(
prec0_expr
, prec1_expr
, prec2_expr
, prec3_expr
, prec4_expr
, not_expr
);

struct indent
{
std::size_t cur;
};
indent operator+(indent lhs, std::size_t rhs)
{
return { lhs.cur + rhs };
}
std::ostream& operator<<(std::ostream& os, indent const& v)
{
for (unsigned i = 0; i < v.cur; ++i) os << ' ';
return os;
}
struct is_simple
{
template <typename T>
bool operator()(T const&) const
{
return std::is_same<T, id>::value || std::is_same<T, value>::value;
}
};
struct printer
{
indent indent_;
void operator()(id const& v)
{
std::cout << v;
}
void operator()(value const& v)
{
std::cout << ''' << v << ''';
}
void operator()(nary_expr const& v)
{
std::cout << '(' << v.op << ' ';
printer p{ indent_ + 2 + v.op.size() };
boost::apply_visitor(p, v.exprs[0]);
for (std::size_t i = 1; i < v.exprs.size(); ++i) {
if (boost::apply_visitor(is_simple{}, v.exprs[i])) {
std::cout << ' ';
}
else {
std::cout << 'n' << p.indent_;
}
boost::apply_visitor(p, v.exprs[i]);
}
std::cout << ')';
}
};
int main()
{
std::string s = "not (xyz='a' or xyz='b' or xyz='c') and abc='s' xor (pqr ='v' and xyz='d')";
expr expr;
auto iter = s.cbegin();
if (phrase_parse(iter, s.cend(), prec4_expr_def, x3::space, expr) && iter == s.cend()) {
boost::apply_visitor(printer{}, expr);
}
return 0;
}

它打印：

(xor (and (not (or (= xyz 'a')
(= xyz 'b')
(= xyz 'c')))
(= abc 's'))
(and (= pqr 'v')
(= xyz 'd')))
相关文章：