我有类似json形式的输入
{
type: "name",
value: <json_value>,
}
类型名称可以是&#34; int1&#34;,&#34; int2a&#34;,&#34; int2b&#34; (当然,我简化了实际情况以提供最小的相关代码)。
该值始终确认为JSON语法,但也取决于类型名称。可能的情况是:
type: int1 => expected value: <number>
type: int2a => expected value: [ <number>, <number> ]
type: int2b => expected value: [ <number>, <number> ]
我需要将输入解析为以下数据类型:
struct int_1 { int i1; };
struct int_2a { int i1, i2; };
struct int_2b { int i1, i2; };
using any_value = boost::variant<int_1, int_2a, int_2b>;
struct data { std::string type; any_value value; };
我将关键字解析器与Nabialek技巧相结合。我创建符号表并将指向 int_1 , int_2a 和 int_2b 解析器的指针存储到其中:
using value_rule_type = qi::rule<It, any_value (), Skipper>;
qi::symbols<char, value_rule_type *> value_selector;
qi::rule<It, int_1 (), Skipper> int1_parser;
qi::rule<It, int_2a (), Skipper> int2a_parser;
qi::rule<It, int_2b (), Skipper> int2b_parser;
value_rule_type int1_rule, int2a_rule, int2b_rule;
int1_parser = int_ ;
int2a_parser = '[' >> int_ >> ',' >> int_ >> ']';
int2b_parser = '[' >> int_ >> ',' >> int_ >> ']';
int1_rule = int1_parser;
int2a_rule = int2a_parser;
int2b_rule = int2b_parser;
value_selector.add
( "\"int1\"", &int1_rule )
("\"int2a\"", &int2a_rule )
("\"int2b\"", &int2b_rule )
;
我使用关键字解析器来解析外部数据结构:
data_
%= eps [ _a = px::val (nullptr) ]
> '{' > (
kwd ( lit ("\"type\"" ) ) [ ':' >> parsed_type_ (_a) >> ',' ]
/ kwd ( lit ("\"value\"") ) [ ':' >> value_ (_a) >> ',' ]
) > '}'
;
parsed_type_ rule在这里查看符号表中的类型名称,并将数据规则的局部变量设置为找到的规则指针。
parsed_type_ %= raw[value_selector [ _r1 = _1 ]];
而且value_规则具有Nabialek技巧的通常形式:
value_ = lazy (*_r1);
这个解析器工作正常(live demo)...除了在类型名称之前传递值的情况:
{
value: <json_value>,
type: "name",
}
因为我们在存储的指针指针中有NULL,而#34;类型的解析器&#34;字段尚未运行,程序在解析&#34;值&#34;期间崩溃。字段。
我想解决这个问题。如果值字段在类型名称字段后面,我喜欢应用当前的解析器逻辑(如在现场演示中)。
如果value在类型键之前,我想用json解析器预先解析value字段(只是为了找到该字段的结束边界)并以迭代器范围的形式存储该字段一些局部变量。解析器得到&#34;键入&#34;字段,我想在存储的范围内启动特定的解析器 - int1_rule,int2a_rule或int2b_rule。
所以,表达式
value_ = lazy (*_r1);
应该改为:
if (_r1 == NULL) {
<parse json value>
<store raw range into local variable>
} else {
lazy (*_r1);
}
和表达
parsed_type_ %= raw[value_selector [ _r1 = _1 ]];
应该扩展为:
if (has stored range) parse it with lazy (*_r1);
不幸的是,我不知道如何实现它,或者根本不可能。
为方便起见,我将stackoverflow上的简化JSON解析器包含到我的现场演示中。
问题:精神上是否有可能?如果是的话,怎么做?
PS。完整的演示源:
#define BOOST_SPIRIT_USE_PHOENIX_V3
#include <boost/optional.hpp>
#include <boost/variant.hpp>
#include <boost/variant/recursive_variant.hpp>
#include <boost/spirit/include/qi.hpp>
#include <boost/spirit/include/phoenix.hpp>
#include <boost/spirit/repository/include/qi_kwd.hpp>
#include <boost/spirit/repository/include/qi_keywords.hpp>
#include <boost/fusion/adapted.hpp>
#include <boost/fusion/adapted/std_pair.hpp>
namespace spirit = ::boost::spirit;
namespace qi = spirit::qi;
namespace px = ::boost::phoenix;
namespace json {
struct null {
constexpr bool operator== (null) const { return true; }
};
template<typename Ch, typename Tr>
std::basic_ostream<Ch, Tr>&
operator<< (std::basic_ostream<Ch, Tr>& os, null) { return os << "null"; }
using text = std::string;
using value = boost::make_recursive_variant<
null
, text // string
, double // number
, std::map<text, boost::recursive_variant_> // object
, std::vector<boost::recursive_variant_> // array
, bool
>::type;
using member = std::pair<text, value>;
using object = std::map<text, value>;
using array = std::vector<value>;
static auto const null_ = "null" >> qi::attr (null {});
static auto const bool_ =
"true" >> qi::attr (true) | "false" >> qi::attr (false);
#if 0
static auto const text_ =
'"' >> qi::raw [*('\\' >> qi::char_ | ~qi::char_('"'))] >> '"';
#endif
template <typename It, typename Skipper = qi::space_type>
struct grammar : qi::grammar<It, value (), Skipper>
{
grammar () : grammar::base_type (value_)
{
using namespace qi;
text_ = '"' >> qi::raw [*('\\' >> qi::char_ | ~qi::char_('"'))] >> '"';
value_ = null_ | bool_ | text_ | double_ | object_ | array_;
member_ = text_ >> ':' >> value_;
object_ = '{' >> -(member_ % ',') >> '}';
array_ = '[' >> -(value_ % ',') >> ']';
BOOST_SPIRIT_DEBUG_NODES((value_)(member_)(object_)(array_))
}
private:
qi::rule<It, std::string ()> text_;
qi::rule<It, json:: value (), Skipper> value_;
qi::rule<It, json::member (), Skipper> member_;
qi::rule<It, json::object (), Skipper> object_;
qi::rule<It, json:: array (), Skipper> array_;
};
template <typename Range,
typename It = typename boost::range_iterator<Range const>::type>
value parse(Range const& input)
{
grammar<It> g;
It first(boost::begin(input)), last(boost::end(input));
value parsed;
bool ok = qi::phrase_parse(first, last, g, qi::space, parsed);
if (ok && (first == last))
return parsed;
throw std::runtime_error("Remaining unparsed: '" +
std::string(first, last) + "'");
}
} // namespace json
namespace mine {
struct int_1 { int_1 (int i) : i1 (i) {} int_1 () : i1 () {} int i1; };
struct int_2a { int i1, i2; };
struct int_2b { int i1, i2; };
using any_value = boost::variant<int_1, int_2a, int_2b>;
struct data { std::string type; any_value value; };
template <class C, class T> std::basic_ostream<C,T>&
operator<< (std::basic_ostream<C,T>& os, int_1 const& i)
{
return os << "{int1:" << i.i1 << '}';
}
template <class C, class T> std::basic_ostream<C,T>&
operator<< (std::basic_ostream<C,T>& os, int_2a const& i)
{
return os << "{int2a:" << i.i1 << ',' << i.i2 << '}';
}
template <class C, class T> std::basic_ostream<C,T>&
operator<< (std::basic_ostream<C,T>& os, int_2b const& i)
{
return os << "{int2b:" << i.i1 << ',' << i.i2 << '}';
}
template <class C, class T> std::basic_ostream<C,T>&
operator<< (std::basic_ostream<C,T>& os, data const& d)
{
return os << "{type=" << d.type << ",value=" << d.value << '}';
}
}
BOOST_FUSION_ADAPT_STRUCT(mine::int_1, (int, i1) )
BOOST_FUSION_ADAPT_STRUCT(mine::int_2a, (int, i1) (int, i2) )
BOOST_FUSION_ADAPT_STRUCT(mine::int_2b, (int, i1) (int, i2) )
BOOST_FUSION_ADAPT_STRUCT(mine::data,(std::string,type)(mine::any_value,value))
namespace mine {
template <typename It, typename Skipper = qi::space_type>
struct grammar: qi::grammar<It, data (), Skipper>
{
grammar () : grammar::base_type (start)
{
using namespace qi;
using spirit::repository::qi::kwd;
int1_parser = int_ ;
int2a_parser = '[' >> int_ >> ',' >> int_ >> ']';
int2b_parser = '[' >> int_ >> ',' >> int_ >> ']';
int1_rule = int1_parser;
int2a_rule = int2a_parser;
int2b_rule = int2b_parser;
value_selector.add
( "\"int1\"", &int1_rule )
("\"int2a\"", &int2a_rule )
("\"int2b\"", &int2b_rule )
;
start = data_.alias ();
parsed_type_ %= raw[value_selector [ _r1 = _1 ]];
value_ = lazy (*_r1);
data_
%= eps [ _a = px::val (nullptr) ]
> '{' > (
kwd ( lit ("\"type\"" ) ) [ ':' >> parsed_type_ (_a) >> ',' ]
/ kwd ( lit ("\"value\"") ) [ ':' >> value_ (_a) >> ',' ]
) > '}'
;
on_error<fail>(start,
px::ref(std::cout)
<< "Error! Expecting "
<< qi::_4
<< " here: '"
<< px::construct<std::string>(qi::_3, qi::_2)
<< "'\n"
);
}
private:
using value_rule_type = qi::rule<It, any_value (), Skipper>;
qi::rule<It, data (), Skipper> start;
qi::rule<It, data (), qi::locals<value_rule_type *>, Skipper> data_;
qi::symbols<char, value_rule_type *> value_selector;
qi::rule<It, int_1 (), Skipper> int1_parser;
qi::rule<It, int_2a (), Skipper> int2a_parser;
qi::rule<It, int_2b (), Skipper> int2b_parser;
value_rule_type int1_rule, int2a_rule, int2b_rule;
qi::rule<It, std::string (value_rule_type *&) > parsed_type_;
qi::rule<It, any_value (value_rule_type *&), Skipper> value_;
};
template <typename Range,
typename It = typename boost::range_iterator<Range const>::type>
data parse(Range const& input)
{
grammar<It> g;
It first(boost::begin(input)), last(boost::end(input));
data parsed;
bool ok = qi::phrase_parse(first, last, g, qi::space, parsed);
if (ok && (first == last))
return parsed;
throw std::runtime_error("Remaining unparsed: '" +
std::string(first, last) + "'");
}
}
static std::string const sample1 = R"(
{
"type": "int1",
"value": 111,
})";
static std::string const sample2 = R"(
{
"type": "int2a",
"value": [ 111, 222 ],
})";
static std::string const sample3 = R"(
{
"type": "int2b",
"value": [ 111, 333 ],
})";
static std::string const sample4 = R"(
{
"value": 111,
"type": "int1",
})";
int main()
{
auto mine = mine::parse(sample1); std::cout << mine << '\n';
mine = mine::parse(sample2); std::cout << mine << '\n';
mine = mine::parse(sample3); std::cout << mine << '\n';
// mine = mine::parse(sample4); std::cout << mine << '\n';
return 0;
}
答案 0 :(得分:1)
精神无法展望未来。
Nabialek技巧与变体替代解析器的目标不同。
因此,从概念上讲,您所描述的内容无法完成:您无法根据将来的类型切换解析器。
Nabialek技巧根本不适合这种情况。您需要的是一般化的JSON数据类型和后处理函数,以便在知道所有必要数据后创建实际的AST鼻子。
我过去发布了全面的JSON语法,并在我自己的项目中亲自使用它们。
您描述的解决方法可能被解释为与此方法类似,但我得到的印象是您努力坚持当前的解析器结构。我说你需要重构解析器(概念上),只有然后你可能会发现你仍然可以重用旧解析器的位。
如果你关心并且有耐心,当我靠近有互联网的电脑时,我可以尝试一下手。