我正在嘗試使用 Boost spirit x3 將樹決議器的結果存盤到遞回結構中。
我有一個很好的語法和一個很好的ast。但我很難理解他們(拒絕)如何聯系。
這是 AST:
#include <boost/spirit/home/x3.hpp>
#include <boost/spirit/home/x3/support/ast/variant.hpp>
#include <boost/fusion/adapted/struct/adapt_struct.hpp>
#include <boost/fusion/include/adapt_struct.hpp>
#include <optional>
namespace ast
{
struct node
{
// are the optional necessary here?
std::optional<std::vector<node>> children;
std::optional<std::string> name;
std::optional<double> length;
};
}
BOOST_FUSION_ADAPT_STRUCT(ast::node, children, name, length)
然后,這是應該合成這些節點的語法:(我添加了一些我認為非終端決議器正在合成的注釋)。
#include <boost/spirit/home/x3.hpp>
#include "ast.h"
#include <boost/fusion/include/vector.hpp>
#include <boost/fusion/include/std_pair.hpp>
namespace parser
{
namespace x3 = boost::spirit::x3;
// synthesize: it should be a simple ast::node, but the grammar says otherwise???
x3::rule<class tree, ast::node> const tree{"tree"};
x3::rule<class branch> branch{"branch"};
// synthesize: std::string
auto name = x3::lexeme[x3::alpha >> *x3::alnum]; // to be improved later
// synthesize: double
auto length = ':' >> x3::double_;
// synthesize: std::optional<std::string>
auto leaf = -name;
// synthesize: std::pair< std::vector<...>, std::optional<std::string> >
auto internal = '(' >> (branch % ',') >> ')' >> -name;
// synthesized attribute: std::variant<node, std::optional<std::string>
auto subtree = internal | leaf;
// synthesize: std::pair<node, std::optional<double>>
auto const tree_def = x3::skip(x3::blank)[subtree >> -length >> ';' >> x3::eoi];
// synthesize: std::pair<node, std::optional<double>>
auto const branch_def = subtree >> -length;
BOOST_SPIRIT_DEFINE(branch, tree);
} // end namespace parser
// What is that for?
auto tree()
{
return parser::tree;
}
盡管多次嘗試“修復”它,但它無法編譯/boost/spirit/home/x3/operator/detail/sequence.hpp:143:9: error: static_assert failed due to requirement 'actual_size <= expected_size' "Size of the passed attribute is bigger than expected."
我的猜測是 AST 與語法不兼容(我沒有tuple<vector<node>, string, double>從語法中看到任何類似簡單的東西)。如果是這樣,我應該將 AST 復雜化還是找到一種簡單的方法來表達語法,或者它會是語意動作的用例嗎?
uj5u.com熱心網友回復:
首先,事物很少是語意動作的用例(Boost Spirit:“語意動作是邪惡的”?)1。
其次,我的直覺說optional對于向量/字串欄位來說可能是不必要的。在我看來,序列容器已經是optional(其中可選的類似于最大大小為 1 的容器)的概括。
第三,據我所知, 似乎我記錯了std::optional支持可能存在局限性。variant支持Transitioning Boost Spirit 決議器從 boost::variant 到 std::variant。
傳播規則
現在,我建議記住兩件事:
始終盡可能將 AST 宣告與您的規則宣告相匹配。特別是,您的 AST 不區分葉節點和內部節點,這會導致混淆,因為您希望這兩個規則“神奇地”與兩者兼容。
具體說明型別強制。
x3::rule這樣做,你可以明確地使用它。我經常使用的 x3 設備之一是as<T>[p](或者有時是它的功能版本:)as<T>(p, name):template <typename T> struct as_type { auto operator[](auto p) const { struct Tag {}; return x3::rule<Tag, T>{"as"} = p; } }; template <typename T> static inline as_type<T> as{};
請注意,您的一些(大多數?)評論行
// synthesize: std::string
// synthesize: double
// synthesize: std::optional<std::string>
// synthesize: std::pair< std::vector<...>, std::optional<std::string> >
// synthesized attribute: std::variant<node, std::optional<std::string>
// synthesize: std::pair<node, std::optional<double>>
// synthesize: std::pair<node, std::optional<double>>
不是很準確。例如alpha >> *alnum具有fusion::deque<char, std::string> >屬性型別。x3::move_to請注意,在將 實體化為實際系結參考時,傳播機制可能能夠消除細微差別。
事實證明確實如此,請參閱下面的簡化,清理
但實際上,這會引起漣漪:leaf不是optional<string>現在等等。
由于您branch沒有定義屬性型別,因此它的屬性型別實際上是x3::unused_type. 這意味著subtree不包含vector<...>任何一個。我不得不檢查,所以做了一個設施:
template <typename ExpectedAttributeType> void check(auto p) {
static_assert(
std::is_same_v<
ExpectedAttributeType,
typename x3::traits::attribute_of<decltype(p), x3::unused_type>::type>);
};
現在我們可以看到:
check<std::string>(name); // assuming a fixed `name` rule that coerces std::string
check<std::vector<std::string>>('(' >> name % ',' >> ')');
但是對于公開??的決議器unused_type:
check<x3::unused_type>(x3::omit[name]);
check<std::vector<x3::unused_type>>('(' >> x3::omit[name] % ',' >> ')'); // FAILS!
check<x3::unused_type>('(' >> x3::omit[name] % ',' >> ')'); // Passes
明白我所說的保持嚴密控制的意思了嗎?漣漪效應導致屬性型別與您的預期截然不同。
此外,pair<node, optional<double>>不會與您的任何 AST 型別兼容(這恰好是當然 ast::node的)。
展示,不說
因為我知道你是一個快速學習者,我不知道如何最好地解釋我一步一步做的轉換,我將只向你展示將上述原則應用于問題代碼的結果:
x3::rule<class branch, ast::node> node{"node"};
auto name = as<std::string>[x3::lexeme[x3::alpha >> *x3::alnum]];
auto length = as<double>[':' >> x3::double_];
auto leaf = as<std::string>[name | x3::attr(std::string{})];
auto children = '(' >> (node % ',') >> ')' | x3::attr(ast::nodes{});
auto node_def = children >> leaf >> -length;
auto tree = x3::skip(x3::blank)[node >> ';' >> x3::eoi];
BOOST_SPIRIT_DEFINE(node);
你可以你的期望:
void checks() {
check<double>(length);
check<std::string>(leaf);
check<std::string>(name);
check<ast::nodes>('(' >> node % ',' >> ')'); // Passes
//check<ast::nodes>(children); // FAILS, actually variant<ast::nodes, ast::nodes> (!!)
check<ast::nodes>(as<ast::nodes>[children]); // Trivially compatible
check<ast::node>(tree);
}
請注意內部機制是多么微妙。說句公道話,靈氣在過去比X3更容易預測和寬容。我懷疑這種差異可能有利于可維護性和編譯時間?_
測驗
從 Grammar 借用一些測驗用例無法使用 Boost Spirit X3 決議樹并除錯列印決議的 AST:
Live On Coliru版畫
============ running internal tests:
"(,)" PASS -> node { [node { [], "" }, node { [], "" }], "" }
"(A,B)F" PASS -> node { [node { [], "A" }, node { [], "B" }], "F" }
"(A:10,B:10)F" PASS -> node { [node { [], "A":10 }, node { [], "B":10 }], "F" }
============ running tree tests:
";" PASS -> node { [], "" }
"(,);" PASS -> node { [node { [], "" }, node { [], "" }], "" }
"(,,(,));" PASS -> node { [node { [], "" }, node { [], "" }, node { [node { [], "" }, node { [], "" }], "" }], "" }
"(A,B,(C,D));" PASS -> node { [node { [], "A" }, node { [], "B" }, node { [node { [], "C" }, node { [], "D" }], "" }], "" }
"(A,B,(C,D)E)F;" PASS -> node { [node { [], "A" }, node { [], "B" }, node { [node { [], "C" }, node { [], "D" }], "E" }], "F" }
"(:0.1,:0.2,(:0.3,:0.4):0.5);" PASS -> node { [node { [], "":0.1 }, node { [], "":0.2 }, node { [node { [], "":0.3 }, node { [], "":0.4 }], "":0.5 }], "" }
"(:0.1,:0.2,(:0.3,:0.4):0.5):0.0;" PASS -> node { [node { [], "":0.1 }, node { [], "":0.2 }, node { [node { [], "":0.3 }, node { [], "":0.4 }], "":0.5 }], "":0 }
"(A:0.1,B:0.2,(C:0.3,D:0.4):0.5);" PASS -> node { [node { [], "A":0.1 }, node { [], "B":0.2 }, node { [node { [], "C":0.3 }, node { [], "D":0.4 }], "":0.5 }], "" }
"(A:0.1,B:0.2,(C:0.3,D:0.4)E:0.5)F;" PASS -> node { [node { [], "A":0.1 }, node { [], "B":0.2 }, node { [node { [], "C":0.3 }, node { [], "D":0.4 }], "E":0.5 }], "F" }
"((B:0.2,(C:0.3,D:0.4)E:0.5)F:0.1)A;" PASS -> node { [node { [node { [], "B":0.2 }, node { [node { [], "C":0.3 }, node { [], "D":0.4 }], "E":0.5 }], "F":0.1 }], "A" }
簡化,清理
全部完成后,我通常會進去并洗掉不必要的規則實體化。
結果,不出所料,確實將 AST 與您的規則 1:1 匹配,這使得一切都非常兼容而無需強制。在您的情況下,我可能還會洗掉optional<double>,所以讓我向您展示,以防您感興趣:
struct node {
nodes children;
std::string name;
double length;
};
規則變成:
x3::rule<class branch, ast::node> node{"node"};
auto name = x3::lexeme[x3::alpha >> *x3::alnum];
auto length = ':' >> x3::double_ | x3::attr(0.0);
auto leaf = name | x3::attr(std::string{});
auto children = '(' >> (node % ',') >> ')' | x3::attr(ast::nodes{});
auto node_def = children >> leaf >> -length;
auto tree = x3::skip(x3::blank)[node >> ';' >> x3::eoi];
BOOST_SPIRIT_DEFINE(node);
為了好的措施,讓我們重新措辭checks以僅驗證預期的兼容性:
template <typename ExpectedAttributeType> void compatible(auto p) {
static_assert(
std::is_same_v<ExpectedAttributeType,
typename x3::traits::attribute_of<
decltype(x3::rule<struct _, ExpectedAttributeType>{} = p),
x3::unused_type>::type>);
};
void checks() {
compatible<double>(length);
compatible<std::string>(leaf);
compatible<std::string>(name);
compatible<ast::nodes>(children);
compatible<ast::node>(tree);
}
完整清單
生活在 Coliru
#include <boost/fusion/include/adapted.hpp>
#include <boost/spirit/home/x3.hpp>
#include <iomanip>
#include <iostream>
namespace ast {
using nodes = std::vector<struct node>;
struct node {
nodes children;
std::string name;
double length;
};
// for debug output
static inline std::ostream& operator<<(std::ostream& os, node const& n) {
os << "node { [";
for (auto sep = ""; auto& c : n.children)
os << std::exchange(sep, ", ") << c;
os << "], " << quoted(n.name) ;
if (n.length != 0)
os << ":" << n.length;
return os << " }";
}
} // namespace ast
BOOST_FUSION_ADAPT_STRUCT(ast::node, children, name, length)
#include <boost/fusion/include/vector.hpp>
#include <boost/fusion/include/std_pair.hpp>
namespace parser {
namespace x3 = boost::spirit::x3;
static x3::rule<class branch, ast::node> const node{"node"};
static auto const name = x3::lexeme[x3::alpha >> *x3::alnum];
static auto const length = ':' >> x3::double_ | x3::attr(0.0);
static auto const leaf = name | x3::attr(std::string{});
static auto const children = '(' >> (node % ',') >> ')' | x3::attr(ast::nodes{});
static auto const node_def = children >> leaf >> -length;
static auto const tree = x3::skip(x3::blank)[node >> ';' >> x3::eoi];
BOOST_SPIRIT_DEFINE(node);
template <typename ExpectedAttributeType> void compatible(auto p) {
static_assert(
std::is_same_v<ExpectedAttributeType,
typename x3::traits::attribute_of<
decltype(x3::rule<struct _, ExpectedAttributeType>{} = p),
x3::unused_type>::type>);
};
void checks() {
compatible<double>(length);
compatible<std::string>(leaf);
compatible<std::string>(name);
compatible<ast::nodes>(children);
compatible<ast::node>(tree);
}
} // namespace parser
namespace test {
void run_tests(auto name, auto p, std::initializer_list<char const*> cases) {
std::cerr << "============ running " << name << " tests:\n";
for (std::string const input : cases)
{
ast::node n;
auto ok = parse(begin(input), end(input), p, n);
std::cout << quoted(input) << "\t " << (ok ? "PASS" : "FAIL");
if (ok)
std::cout << " -> " << n << std::endl;
else
std::cout << std::endl;
}
}
void internal() {
run_tests("internal", parser::node,
{
"(,)",
"(A,B)F",
"(A:10,B:10)F",
});
}
void tree() {
run_tests("tree", parser::tree,
{
";",
"(,);",
"(,,(,));",
"(A,B,(C,D));",
"(A,B,(C,D)E)F;",
"(:0.1,:0.2,(:0.3,:0.4):0.5);",
"(:0.1,:0.2,(:0.3,:0.4):0.5):0.0;",
"(A:0.1,B:0.2,(C:0.3,D:0.4):0.5);",
"(A:0.1,B:0.2,(C:0.3,D:0.4)E:0.5)F;",
"((B:0.2,(C:0.3,D:0.4)E:0.5)F:0.1)A;",
});
}
} // namespace test
int main() {
test::internal();
test::tree();
}
0.0具有相同的輸出(除錯輸出中的抑制長度除外)。
1 公平地說,在 X3 中,我會更樂意考慮它,因為撰寫它們變得更加自然。
轉載請註明出處,本文鏈接:https://www.uj5u.com/caozuo/537780.html
標籤:C 解析促进振奋精神
上一篇:我將如何使用Javascript中的正則運算式決議特定的JSON引數?
下一篇:BOOST_FUSION_ADAPT_STRUCT使用帶有std::vector<self_type>成員的遞回結構
