提升精神气：用元组向量绑定到结构

Question

Boost Spirit Qi 解析当然是 C++ 的独特应用，它具有陡峭的学习曲线。在这种情况下，我试图解析一个字符串，其中包含 struct 的语法正确的 C++ 列表初始化，其中 std::vector 为 std::tuple<std::string, short>。这是 struct:

的声明

typedef std::vector<std::tuple<std::string, int>> label_t;

struct BulkDataParmas
{
    std::string strUUID;
    short subcam;
    long long pts_beg;
    long long pts_len;
    long long pts_gap;
    label_t labels;
};

这是我尝试将这种结构绑定到 Qi 属性的失败尝试。如果我也注释掉 struct 的 vector 成员，注释掉的 start 会按预期工作。（我也试过 std::pair 而不是 std::tuple）。

BOOST_FUSION_ADAPT_STRUCT
(
    BulkDataParmas,
    (std::string, strUUID)
    (short, subcam)
    (long long, pts_beg)
    (long long, pts_len)
    (long long, pts_gap)
    (label_t, labels)
)



template <typename Iterator>
struct load_parser : boost::spirit::qi::grammar<Iterator, BulkDataParmas(), boost::spirit::ascii::space_type>
{
    load_parser() : load_parser::base_type(start)
    {
        namespace qi = boost::spirit::qi;
        namespace ascii = boost::spirit::ascii;
        using qi::attr;
        using qi::short_;
        using qi::int_;
        using qi::long_long;
        using qi::lit;
        using qi::xdigit;
        using qi::lexeme;
        using ascii::char_;
        using boost::proto::deep_copy;

        auto hex2_ = deep_copy(xdigit >> xdigit >> xdigit >> xdigit);
        auto hex4_ = deep_copy(hex2_ >> hex2_);
        auto hex6_ = deep_copy(hex4_ >> hex2_);
        auto fmt_  = deep_copy('"' >> hex4_ >> char_('-') >> hex2_ >> char_('-') >> hex2_ >> char_('-') >> hex2_ >> char_('-') >> hex6_ >> '"');
        uuid = qi::as_string[fmt_];

        quoted_string %= lexeme['"' >> +(char_ - '"') >> '"'];

        label = '{' >> quoted_string >> ',' >> int_ >> '}';

        start = '{' >>  uuid >> ',' >> short_ >> ',' >> long_long >> ',' >> long_long >> ',' >> long_long >> ',' >> '{' >> -(label >> *(',' >> label)) >>'}' >> '}';
//        start = '{' >>  uuid >> ',' >> short_ >> ',' >> long_long >> ',' >> long_long >> ',' >> long_long >> '}';
    }

private:

    boost::spirit::qi::rule<Iterator, std::string()> uuid;
    boost::spirit::qi::rule<Iterator, std::string()> quoted_string;
    boost::spirit::qi::rule<Iterator, std::string(), boost::spirit::ascii::space_type> label;
    boost::spirit::qi::rule<Iterator, BulkDataParmas(), boost::spirit::ascii::space_type> start;
};

这是一个要解析的示例字符串：

"{ \"68965363-2d87-46d4-b05d-f293f2c8403b\", 0, 1583798400000000, 86400000000, 600000000, { { \"motorbike\", 5 }, { \"aeroplane\", 6 } } };"

Answer 1

我正在回答我自己的问题。我犯了两个错误。首先，规则 label 的属性类型错误，std::string() 而不是 std::tuple<std::string, int>()。

第二个错误是我需要#include <boost/fusion/adapted/std_tuple.hpp>。我只是偶然发现的，因为这不在 Spirit 2.5 文档中。

template <typename Iterator>
struct load_parser : boost::spirit::qi::grammar<Iterator, BulkDataParmas(), boost::spirit::ascii::space_type>
{
    load_parser() : load_parser::base_type(start)
    {
        namespace qi = boost::spirit::qi;
        namespace ascii = boost::spirit::ascii;
        using qi::attr;
        using qi::short_;
        using qi::int_;
        using qi::long_long;
        using qi::lit;
        using qi::xdigit;
        using qi::lexeme;
        using ascii::char_;
        using boost::proto::deep_copy;

        auto hex2_ = deep_copy(xdigit >> xdigit >> xdigit >> xdigit);
        auto hex4_ = deep_copy(hex2_ >> hex2_);
        auto hex6_ = deep_copy(hex4_ >> hex2_);
        auto fmt_  = deep_copy('"' >> hex4_ >> char_('-') >> hex2_ >> char_('-') >> hex2_ >> char_('-') >> hex2_ >> char_('-') >> hex6_ >> '"');
        uuid = qi::as_string[fmt_];

        quoted_string %= lexeme['"' >> +(char_ - '"') >> '"'];

        label = '{' >> quoted_string >> ',' >> int_ >> '}';

        start = '{' >>  uuid >> ',' >> short_ >> ',' >> long_long >> ',' >> long_long >> ',' >> long_long >> ',' >> '{' >> -(label >> *(',' >> label)) >>'}' >> '}';
//        start = '{' >>  uuid >> ',' >> short_ >> ',' >> long_long >> ',' >> long_long >> ',' >> long_long >> '}';
    }

private:

    boost::spirit::qi::rule<Iterator, std::string()> uuid;
    boost::spirit::qi::rule<Iterator, std::string()> quoted_string;
    boost::spirit::qi::rule<Iterator, std::tuple<std::string, int>(), boost::spirit::ascii::space_type> label;
    boost::spirit::qi::rule<Iterator, BulkDataParmas(), boost::spirit::ascii::space_type> start;
};

测试代码：

void doTestParser2()
{
    for
    (
        auto& input : std::list<std::string>
        {
            "{ \"68965363-2d87-46d4-b05d-f293f2c8403b\", 0, 1583798400000000, 86400000000, 600000000, {  } };",
            "{ \"68965363-2d87-46d4-b05d-f293f2c8403b\", 0, 1583798400000000, 86400000000, 600000000, { { \"motorbike\", 5 } } };",
            "{ \"68965363-2d87-46d4-b05d-f293f2c8403b\", 0, 1583798400000000, 86400000000, 600000000, { { \"motorbike\", 5 }, { \"aeroplane\", 6 } } };"
        }
    )
    {
        using namespace boost::spirit;

        auto f(std::begin(input)), l(std::end(input));
        load_parser<decltype(f)> p;

        try
        {
            BulkDataParmas result { };
            std::string sresult { };
            bool ok = qi::phrase_parse(f, l, p > ';', qi::ascii::space, result);

            if (!ok)
                std::cerr << "invalid input" << std::endl;
            else
            {
                std::cout << "ok: " << input << std::endl;
                std::cout << "UUID:     " << result.strUUID << std::endl;
                std::cout << "subcam:   " << result.subcam << std::endl;
                std::cout << "pts_beg:  " << result.pts_beg << std::endl;
                std::cout << "pts_len:  " << result.pts_len << std::endl;
                std::cout << "pts_gap:  " << result.pts_gap << std::endl;
                for (auto const& tup : result.labels)
                {
                    std::cout << "label:    " << std::get<0>(tup) << std::endl;
                    std::cout << "level:    " << std::get<1>(tup) << std::endl;
                }

            }

        }
        catch (const qi::expectation_failure<decltype(f)>& e)
        {
            std::cerr << "expectation_failure at '" << std::string(e.first, e.last) << "'\n";
        }
    }
}

Answer 2

除了你提到的两件事（这是正确的），我建议

一些简化：

uuid = '"' >> qi::raw [
    hex_<4>{} >> qi::repeat(3)['-' >> hex_<2>{}] >> '-' >> hex_<6>{}
] >> '"';

注意，这将删除所有子表达式，as-string 和 deepcopy，而不是使用整数解析器：

template<int N> using hex_ = boost::spirit::qi::int_parser<std::intmax_t, 16, 2*N, 2*N>;

raw[] 解析器将很好地公开匹配的源字符串。

接下来，
```
quoted_string = '"' >> *~qi::char_('"') >> '"';
```
这里我建议使用 * 接受空字符串（这经常 "the point" 个带引号的字符串，因此我们可以明确嵌入空格或故意为空的字符串）。另外，使用 ~charset 是更有效率。

也删除了 lexeme[] 因为规则已经在没有船长的情况下声明了。

完成：

label = '{' >> quoted_string >> ',' >> qi::int_ >> '}';

start = qi::skip(ascii::space) [ '{'
    >> uuid      >> ','
    >> qi::auto_ >> ','
    >> qi::auto_ >> ','
    >> qi::auto_ >> ','
    >> qi::auto_ >> ','
    >> '{' >> -(label % ',') >> '}'
    >> '}' >> ';'
];

请注意，我合并了船长的选择。所以你不必乏味地在 phrase_parse 中传递正确的东西。船长通常不是调用者应该能够改变的东西。

现在让我们也对改编进行现代化改造：
```
BOOST_FUSION_ADAPT_STRUCT(BulkDataParams, strUUID, subcam, pts_beg, pts_len, pts_gap, labels)
```
之后您可以以现代方式重新拼写类型，而不会出现任何兼容性问题。请注意，这也是在开始规则中更喜欢 qi::auto_ 的原因，因此您不会在例如解析器结果以预期的方式隐式转换为目标类型。
```
struct BulkDataParams {
    std::string strUUID;
    int16_t subcam;
    int64_t pts_beg;
    int64_t pts_len;
    int64_t pts_gap;
    label_t labels;
};
```

现在让我们输入调试输出和测试主体：

Live On Wandbox

#define BOOST_SPIRIT_DEBUG
#include <boost/spirit/include/qi.hpp>
#include <boost/fusion/adapted/std_tuple.hpp>
#include <iostream>
#include <iomanip>

using label_t = std::vector<std::tuple<std::string, int>>;

namespace std {
    std::ostream& operator<<(std::ostream& os, label_t::value_type const& t) {
        auto const& [k,v] = t;
        return os << "[" << std::quoted(k) << "," << v << "]";
    }

    std::ostream& operator<<(std::ostream& os, label_t const& m) {
        os << "{";
        for (auto&& el:m) os << el << ",";
        return os << "}";
    }
}

struct BulkDataParams {
    std::string strUUID;
    int16_t subcam;
    int64_t pts_beg;
    int64_t pts_len;
    int64_t pts_gap;
    label_t labels;
};

BOOST_FUSION_ADAPT_STRUCT(BulkDataParams, strUUID, subcam, pts_beg, pts_len, pts_gap, labels)

template <typename Iterator> struct load_parser : boost::spirit::qi::grammar<Iterator, BulkDataParams()> {
    load_parser() : load_parser::base_type(start) {
        namespace qi = boost::spirit::qi;
        namespace ascii = boost::spirit::ascii;

        uuid = '"' >> qi::raw [
            hex_<4>{} >> qi::repeat(3)['-' >> hex_<2>{}] >> '-' >> hex_<6>{}
        ] >> '"';

        quoted_string = '"' >> *~qi::char_('"') >> '"';

        label = '{' >> quoted_string >> ',' >> qi::int_ >> '}';

        start = qi::skip(ascii::space) [ '{'
            >> uuid      >> ','
            >> qi::auto_ >> ','
            >> qi::auto_ >> ','
            >> qi::auto_ >> ','
            >> qi::auto_ >> ','
            >> '{' >> -(label % ',') >> '}'
            >> '}' >> ';'
        ];

        BOOST_SPIRIT_DEBUG_NODES(
            (uuid) (quoted_string) (label) (start)
        )
    }

    template<int N> using hex_ = boost::spirit::qi::int_parser<std::intmax_t, 16, 2*N, 2*N>;

  private:
    boost::spirit::qi::rule<Iterator, std::string()> uuid;
    boost::spirit::qi::rule<Iterator, std::string()> quoted_string;
    boost::spirit::qi::rule<Iterator, label_t::value_type(), boost::spirit::ascii::space_type> label;
    boost::spirit::qi::rule<Iterator, BulkDataParams()> start;
};

int main() {

    for (std::string const input : {
        R"({ "68965363-2d87-46d4-b05d-f293f2c8403b", 0, 1583798400000000, 86400000000, 600000000, { { "motorbike", 5 }, { "aeroplane", 6 } } };)",
    })
    {
        auto f = begin(input), l = end(input);
        BulkDataParams bdp;
        load_parser<std::string::const_iterator> p;
        if (parse(f, l, p, bdp)) {
            std::cout << "Parsed: " << boost::fusion::as_vector(bdp) << "\n";
        } else {
            std::cout << "Parse Failed\n";
        }

        if (f != l) {
            std::cout << "Remaining unparsed: " << std::quoted(std::string(f,l)) << "\n";
        }
    }
}

常规输出：

Parsed: (68965363-2d87-46d4-b05d-f293f2c8403b 0 1583798400000000 86400000000 600000000 {["motorbike",5],["aeroplane",6],})

调试输出：

<start>
  <try>{ "68965363-2d87-46d</try>
  <uuid>
    <try>"68965363-2d87-46d4-</try>
    <success>, 0, 158379840000000</success>
    <attributes>[[6, 8, 9, 6, 5, 3, 6, 3, -, 2, d, 8, 7, -, 4, 6, d, 4, -, b, 0, 5, d, -, f, 2, 9, 3, f, 2, c, 8, 4, 0, 3, b]]</attributes>
  </uuid>
  <label>
    <try> { "motorbike", 5 },</try>
    <quoted_string>
      <try>"motorbike", 5 }, { </try>
      <success>, 5 }, { "aeroplane"</success>
      <attributes>[[m, o, t, o, r, b, i, k, e]]</attributes>
    </quoted_string>
    <success>, { "aeroplane", 6 }</success>
    <attributes>[[[m, o, t, o, r, b, i, k, e], 5]]</attributes>
  </label>
  <label>
    <try> { "aeroplane", 6 } </try>
    <quoted_string>
      <try>"aeroplane", 6 } } }</try>
      <success>, 6 } } };</success>
      <attributes>[[a, e, r, o, p, l, a, n, e]]</attributes>
    </quoted_string>
    <success> } };</success>
    <attributes>[[[a, e, r, o, p, l, a, n, e], 6]]</attributes>
  </label>
  <success></success>
  <attributes>[[[6, 8, 9, 6, 5, 3, 6, 3, -, 2, d, 8, 7, -, 4, 6, d, 4, -, b, 0, 5, d, -, f, 2, 9, 3, f, 2, c, 8, 4, 0, 3, b], 0, 1583798400000000, 86400000000, 600000000, [[[m, o, t, o, r, b, i, k, e], 5], [[a, e, r, o, p, l, a, n, e], 6]]]]</attributes>
</start>

提升精神气：用元组向量绑定到结构

Boost Spirit Qi: binding to struct with vector of tuples

c++

parsing

boost

boost-spirit

boost-spirit-qi