I'm very new to peg and pegtl, so probably I'm missing something. I have a grammar very similar to the following one:
using namespace tao::pegtl;
struct A : one<'A'> { };
struct B : one<'B'> { };
struct comp : seq<plus<sor<seq<A, B>, A>>,eof> { };
template< typename Rule >
struct test_action : nothing< Rule > {};
template<>
struct test_action<A>
{
template< typename Input >
static void apply(const Input& in)
{
std::cout << "A";
}
};
template<>
struct test_action<B>
{
template< typename Input >
static void apply(const Input& in)
{
std::cout << "B";
}
};
void test()
{
parse< comp, test_action >(memory_input("AAB", ""));
}
The parse works great, but there is too many activation of test_action::apply. The program outputs "AAAB", because, if I understand well, the parse tries the first alternative (AB) for the first character and fails, then proceeds with the other (A). But even if it "rewinds", it always call test_action::apply. What is the correct way to handle this situation? My intent is to output "AAB", possibly without complicate the grammar.
I asked to pegtl library authors and they kindly give me the correct way: the best thing to do is make your parser construct a parse tree, which is easy to fix when it backtracks using simple push and pop operations.
I developed the code below for who had similar doubts.
avoid backtracking in rules with attached actions:
using namespace tao::pegtl;
struct A : one<'A'> { };
struct B : one<'B'> { };
struct real_A : A {};
struct real_AB : seq<A, B> {};
struct comp : seq<plus<sor<real_AB, real_A>>,eof> { };
template< typename Rule >
struct test_action : nothing< Rule > {};
template<>
struct test_action<real_A>
{
template< typename Input >
static void apply(const Input& in)
{
std::cout << "A";
}
};
template<>
struct test_action<real_AB>
{
template< typename Input >
static void apply(const Input& in)
{
std::cout << "AB";
}
};
void test()
{
parse< comp, test_action >(memory_input("AAB", ""));
}
build a parse tree:
using namespace tao::pegtl;
struct A : one<'A'> { };
struct B : one<'B'> { };
struct comp : seq<plus<sor<seq<A, B>, A>>, eof> { };
template< typename Rule >
struct test_action : nothing< Rule > {};
void test()
{
auto root = parse_tree::parse<comp>(memory_input("AAB", ""));
}