boost::qvm not usable in boost::spirit::x3 #720
Comments
Tip: use clang++ because of g++ 'internal compiler error' |
|
Try mimicking for qvm what's done with boost::array in fusion here:
https://www.boost.org/doc/libs/develop/libs/fusion/doc/html/fusion/adapted/boost__array.html
It works for me:
… //copy&pasted from:
// ./qvm-not-usable.1rule-deque.cpp
//then just changed deque to boost::array & included the fusion
adaptation for boost::array.
//===================
#include <boost/spirit/home/x3.hpp>
#include <boost/fusion/adapted/boost_array.hpp>
namespace x3 = boost::spirit::x3;
namespace fusion = boost::fusion;
using scalar_t = int;
auto const scalar =
x3::int_;
#define VECTOR_N 3
int constexpr vector_n=VECTOR_N;
#include <boost/array.hpp>
using vector_t =
boost::array<scalar_t,vector_n>
;
#include <boost/fusion/adapted/boost_array.hpp>
#include <boost/fusion/include/io.hpp>
using fusion::operators::operator<<;
auto const vector_def
= x3::lit("center")
> x3::lit('=')
> x3::lit('<')
> scalar
> x3::lit(',')
> scalar
> x3::lit(',')
> scalar
> x3::lit('>')
> x3::eoi
;
auto const vector = x3::rule<struct vector_, vector_t>{"vector"}
= vector_def
;
#include <iostream>
int main() {
std::string input = "center = <1,2,3>";
vector_t result;
auto first=input.begin();
auto last=input.end();
x3::phrase_parse(first, last, vector, x3::space, result);
std::cout<<"result="<<result<<";\n";
return 0;
}
On 3/27/2022 4:41 PM, Mike Gresens wrote:
|#include <boost/qvm/vec.hpp> #include <boost/spirit/home/x3.hpp>
namespace qvm = boost::qvm; namespace x3 = boost::spirit::x3; using
scalar_t = float; auto const scalar = x3::rule<struct scalar_,
scalar_t>{"scalar"} = x3::float_; using vector_t = qvm::vec<scalar_t,
3>; template <size_t N> struct foo { template <typename C> void
operator()(C const& ctx) const { x3::_val(ctx).a[N] = x3::_attr(ctx);
} }; auto const vector = x3::rule<struct vector_, vector_t>{"vector"}
= x3::lit('<') > scalar [foo<0>()] > x3::lit(',') > scalar [foo<1>()]
> x3::lit(',') > scalar [foo<2>()] > x3::lit('>'); int main() {
std::string input = "center = <0,0,0>"; vector_t result;
x3::phrase_parse(input.begin(), input.end(), x3::lit("center") >
x3::lit('=') > vector > x3::eoi, x3::space, result); return 0; } |
Tip: use clang++ because of g++ 'internal compiler error'
—
Reply to this email directly, view it on GitHub
<#720 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAEFBA5AIADUABFSGQGUU7TVCDPY5ANCNFSM5RYXHM2Q>.
You are receiving this because you are subscribed to this
thread.Message ID: ***@***.***>
|
|
Actually, there not even a need for any rule's in this case. This works
for me:
x3::phrase_parse(first, last, vector_def, x3::space, result);
and outputs:
result=(1 2 3);
…On 3/29/2022 7:27 AM, cppljevans wrote:
Try mimicking for qvm what's done with boost::array in fusion here:
https://www.boost.org/doc/libs/develop/libs/fusion/doc/html/fusion/adapted/boost__array.html
It works for me:
> //copy&pasted from:
> // ./qvm-not-usable.1rule-deque.cpp
> //then just changed deque to boost::array & included the fusion
> adaptation for boost::array.
> //===================
> #include <boost/spirit/home/x3.hpp>
> #include <boost/fusion/adapted/boost_array.hpp>
>
> namespace x3 = boost::spirit::x3;
> namespace fusion = boost::fusion;
>
> using scalar_t = int;
>
> auto const scalar =
> x3::int_;
> #define VECTOR_N 3
> int constexpr vector_n=VECTOR_N;
> #include <boost/array.hpp>
> using vector_t =
> boost::array<scalar_t,vector_n>
> ;
> #include <boost/fusion/adapted/boost_array.hpp>
> #include <boost/fusion/include/io.hpp>
> using fusion::operators::operator<<;
>
> auto const vector_def
> = x3::lit("center")
> > x3::lit('=')
> > x3::lit('<')
> > scalar
> > x3::lit(',')
> > scalar
> > x3::lit(',')
> > scalar
> > x3::lit('>')
> > x3::eoi
> ;
>
> auto const vector = x3::rule<struct vector_, vector_t>{"vector"}
> = vector_def
> ;
>
> #include <iostream>
> int main() {
> std::string input = "center = <1,2,3>";
> vector_t result;
> auto first=input.begin();
> auto last=input.end();
> x3::phrase_parse(first, last, vector, x3::space, result);
> std::cout<<"result="<<result<<";\n";
> return 0;
> }
On 3/27/2022 4:41 PM, Mike Gresens wrote:
> |#include <boost/qvm/vec.hpp> #include <boost/spirit/home/x3.hpp>
> namespace qvm = boost::qvm; namespace x3 = boost::spirit::x3; using
> scalar_t = float; auto const scalar = x3::rule<struct scalar_,
> scalar_t>{"scalar"} = x3::float_; using vector_t = qvm::vec<scalar_t,
> 3>; template <size_t N> struct foo { template <typename C> void
> operator()(C const& ctx) const { x3::_val(ctx).a[N] = x3::_attr(ctx);
> } }; auto const vector = x3::rule<struct vector_, vector_t>{"vector"}
> = x3::lit('<') > scalar [foo<0>()] > x3::lit(',') > scalar [foo<1>()]
> > x3::lit(',') > scalar [foo<2>()] > x3::lit('>'); int main() {
> std::string input = "center = <0,0,0>"; vector_t result;
> x3::phrase_parse(input.begin(), input.end(), x3::lit("center") >
> x3::lit('=') > vector > x3::eoi, x3::space, result); return 0; } |
>
> Tip: use clang++ because of g++ 'internal compiler error'
>
> —
> Reply to this email directly, view it on GitHub
>
<#720 (comment)>,
> or unsubscribe
>
<https://github.com/notifications/unsubscribe-auth/AAEFBA5AIADUABFSGQGUU7TVCDPY5ANCNFSM5RYXHM2Q>.
> You are receiving this because you are subscribed to this
> thread.Message ID: ***@***.***>
>
—
Reply to this email directly, view it on GitHub
<#720 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAEFBA7FUNWH5MAUUIXVHXDVCLZMPANCNFSM5RYXHM2Q>.
You are receiving this because you are subscribed to this
thread.Message ID: ***@***.***>
|
|
Thanks for your answers @cppljevans But your parser produces boost::array and not boost::qvm::vec Parsing as array is simple... But the goal is to produce a boost::qvm::vec |
|
But my point was that you could just duplicate the
fusion/adapted/boost_array.hpp file and just replace boost::array with
qvm::vec
and rename to adapted_qvm.hpp or something similar.
…On 3/29/2022 10:09 AM, Mike Gresens wrote:
Thanks for your answers @cppljevans <https://github.com/cppljevans>
But your parser produces boost::array and not boost::qvm::vec
Parsing as array is simple...
just use
#include <boost/fusion/adapted/std_array.hpp>
But the goal is to produce a boost::qvm::vec
—
Reply to this email directly, view it on GitHub
<#720 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAEFBA5YSG4KH5U4TRDP4UDVCMML3ANCNFSM5RYXHM2Q>.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
|
@octopus-prime I've started to create a boost/fusion/adapted/qvm_vec directory which is essentially a copy of boost/fusion/adapted/std_array with "semi-obvious" modifications to adapt it to qvm::vec. It's tedious, but before I go any further, I was wondering if you'd already done that; hence, such efforts on my part would be needless. Please let us know if you are trying to do that, and if so, please let us know the progress. TIA. |
|
Hi @cppljevans no - i did not start any adaptation. Way to much! I am just a qvm/spirit user. So i am very glad to hear that you want to do the adaptation :-) |
|
Without your adaptation my plan would be
It's a lot of boilerplate but should work fine... |
|
Hi, |
|
@octopus-prime , you could use BOOST_FUSION_ADAPT_STRUCT if you're careful to use a 2 element preprocessor sequence, as shown below: //copy&pasted from:
// ./qvm-not-usable.1rule.cpp
//then added adapt_struct with 2 element arg (type,name)
//===================
#include <boost/spirit/home/x3.hpp>
#include <boost/fusion/include/adapt_struct.hpp>
#include <boost/fusion/include/at_c.hpp>
#include <boost/qvm/vec.hpp>
namespace qvm = boost::qvm;
namespace x3 = boost::spirit::x3;
namespace fusion = boost::fusion;
using scalar_t = int;
auto const scalar =
x3::int_;
#define VECTOR_N 2
int constexpr vector_n=VECTOR_N;
using vector_t =
qvm::vec<scalar_t,vector_n>
;
BOOST_FUSION_ADAPT_STRUCT(vector_t, (scalar_t,a[0]), (scalar_t,a[1]));
#include <boost/fusion/include/io.hpp>
using fusion::operators::operator<<;
auto const vector_def
= x3::lit("center")
> x3::lit('=')
> x3::lit('<')
> scalar
> x3::lit(',')
> scalar
> x3::lit('>')
> x3::eoi
;
#include <iostream>
int main() {
std::string input = "center = <1,2>";
vector_t result;
auto first=input.begin();
auto last=input.end();
x3::phrase_parse(first, last, vector_def, x3::space, result);
std::cout<<"result="<<result<<";\n";
return 0;
}which, at least for me, produces output: Compilation finished at Mon Apr 4 06:34:11 |
|
Thank you @lakeweb and @cppljevans ! Looks like you both avoided the critical Instead you see someting like So i tried without success. Still ugly expectation failures. |
@octopus-prime , the following minor change to previous code, when defined(USE_VECTOR_RULE), does use rule: //copy&pasted from:
// ./qvm-not-usable.1rule.cpp
//then added adapt_struct with 2 element arg (type,name)
//===================
#include <boost/spirit/home/x3.hpp>
#include <boost/fusion/include/adapt_struct.hpp>
#include <boost/fusion/include/at_c.hpp>
#include <boost/qvm/vec.hpp>
namespace qvm = boost::qvm;
namespace x3 = boost::spirit::x3;
namespace fusion = boost::fusion;
using scalar_t = int;
#define VECTOR_N 2
int constexpr vector_n=VECTOR_N;
using vector_t =
qvm::vec<scalar_t,vector_n>
;
BOOST_FUSION_ADAPT_STRUCT(vector_t, (scalar_t,a[0]), (scalar_t,a[1]));
#include <boost/fusion/include/io.hpp>
using fusion::operators::operator<<;
auto const scalar
= x3::int_
;
auto const vector_def
= x3::lit("center")
> x3::lit('=')
> x3::lit('<')
> scalar
> x3::lit(',')
> scalar
> x3::lit('>')
> x3::eoi
;
#define USE_VECTOR_RULE
#ifdef USE_VECTOR_RULE
auto const vector_ = x3::rule<struct vector_, vector_t>{"vector"} =
vector_def
;
#endif//USE_VECTOR_RULE
#include <iostream>
int main() {
#ifdef USE_VECTOR_RULE
{//parse error
std::string input = "center = <1>";
vector_t result;
auto first=input.begin();
auto last=input.end();
x3::phrase_parse(first, last, vector_, x3::space, result);
std::cout<<"result="<<result<<";\n";
}
#else
{//parse success
std::string input = "center = <1,2>";
vector_t result;
auto first=input.begin();
auto last=input.end();
x3::phrase_parse(first, last, vector_def, x3::space, result);
std::cout<<"result="<<result<<";\n";
}
#endif
return 0;
}and when run, with input containing error, produces HUGE error output. Is this kind of output what you were hoping for? |
|
These are the compiler erros i opened this issue for... And if you want to handle |
|
This ambiguous conversion error is eliminated if: boost/qvm/vec.hpp is changed to contain: #include <boost/spirit/home/x3/support/unused.hpp>
#include <type_traits>
namespace boost { namespace qvm {
template <class T,int D>
struct
vec
{
T a[D];
template
< class R
, typename Enable=
std::enable_if_t
< !std::is_same_v<R,boost::spirit::x3::unused_type>
>//don't allow conversion to boost::spirit::x3::unused_type
//let unused_type to handle that.
>
operator R() const
{
R r;
assign(r,*this);
return r;
}
};
.
.
.
}}//boost::qvmIOW, the "convenient" conversion operators lead to this "annoyingly inconvenient" error. |
|
OTOH, instead of #720 (comment), To be more specific, here's an example of the change in one primitive where, obviously, in the above, /home/evansl/prog_dev/boost.org/boost.replacements Similar changes could be made to other primitive parsers. In short, by simply adding another template parameter, IgnoreContext, to the parse @djowel , that seems a simple solution. Is there some downside to such a |
Pardon, I've not been actively following this thread. Does this mean changing all primitive parsers? |
|
I think so. The problem is that the primitive parsers have
'unused_type' as the type of one of the arguments,
and unused_type has conversion operator from anything to unused_type.
Unfortunately, qvm::vec also
has a conversion operator that can convert qvm::vec to unused_type.
This leads to the compiler error
about ambiguous conversions. The easiest workaround is as I described
in my previous comment.
…On 4/25/2022 6:11 PM, Joel de Guzman wrote:
@djowel <https://github.com/djowel> , that seems a simple
solution. Is there some downside to such a
solution?
Pardon, I've not been actively following this thread. Does this mean
changing all primitive parsers?
—
Reply to this email directly, view it on GitHub
<#720 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAEFBAZVST6CLWTSSLO76VDVG4RATANCNFSM5RYXHM2Q>.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
|
Isn't this an issue of |
|
The qvm team was working on this issue -> boostorg/qvm#39 |
|
Thanks @octopus-prime ! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
If you try to parse
boost::qvm::vec<float, 3>you fail with compiler errors:
error: conversion from 'boost::qvm::vec<float, 3>' to 'boost::spirit::x3::unused_type' is ambiguousreason:
The text was updated successfully, but these errors were encountered: