Compilation error with g++ about geometry

[awulkiew]

Ok, so you adapt Eigen::Matrix to Boost.Geometry Point concept. The code above looks ok.

What's the program you're compiling? I see that you're creating R-tree and passing std::pair<Eigen::Matrix<...>, double> as ValueType. Is that correct? Or are you calling some algorithm?

The NOT_VALID_INDEXABLE_TYPE MPL assertion occurs when tag<ValueType>::type is neither point_tag, box_tag nor segment_tag. Could you check in your code if the tag is set properly? E.g.: std::cout << boost::is_same<bg::tag<Eigen::Matrix<...>>::type, point_tag>::value;.

How are you including Boost.Geometry in your code and what is the order of includes? It's possible that the order of #includes is important for GCC. This wouldn't be the first time something like this happen with this compiler.

[thk686]

// [[Rcpp::export]]
void test_tag()
{
  Rcout << boost::is_same<bg::tag<Eigen::Matrix<double, 2, 1>>::type, bg::point_tag>::value << std::endl;
}

Returns 1.

[thk686]

Tried moving header to first and last in source file without luck.

[awulkiew]

This may be a problem with wrong traits specializations. If Eigen::Matrix rows, cols, options etc. are int, not std::size_t then DimensionCount in specializations should be of type int too.

Btw, I see that Eigen::Matrix is defined as:

Matrix<typename Scalar,
   int RowsAtCompileTime,
   int ColsAtCompileTime,
   int Options = 0,
   int MaxRowsAtCompileTime = RowsAtCompileTime,
   int MaxColsAtCompileTime = ColsAtCompileTime>

Maybe this particular compiler would prefer to have all 6 template parameters in specializations? You pass only 3.

Ok, so bg::tag seems to work. How about some simple algorithm? Does distance(p1, p2) or disjoint(p1, p2) compile and generate correct result?

[thk686]

It looks like changing the Eigen template types from size_t to int fixes the problem. THK

…

On Tue, Dec 5, 2017 at 7:09 AM, Adam Wulkiewicz ***@***.***> wrote: This may be a problem with wrong traits specializations. If Eigen::Matrix rows, cols, options etc. are int, not std::size_t then DimensionCount in specializations should be of type int too. Btw, I see that Eigen::Matrix is defined as: Matrix<typename Scalar, int RowsAtCompileTime, int ColsAtCompileTime, int Options = 0, int MaxRowsAtCompileTime = RowsAtCompileTime, int MaxColsAtCompileTime = ColsAtCompileTime> Maybe this particular compiler would prefer to have all 6 template parameters in specializations? You pass only 3. Ok, so bg::tag seems to work. How about some simple algorithm? Does distance(p1, p2) or disjoint(p1, p2) compile and generate correct result? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#416 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAkLyXcS9j9GpitNVupwsbbOlMQqH_PAks5s9UBsgaJpZM4O67mU> .

[awulkiew]

Ok, great.