Error when binding template class with inline constructors definition

[farinaleo]

Describe the issue

Hi,
I'm working on a school project and I got an error when I tried to bind a template class with constructors defined with inline. But I don't know why...

Warning: (SrcmlcppIgnoreElement) A cpp element of type "constructor" was stored as CppUnprocessed. Details follow
Warning: (SrcmlcppIgnoreElement) unhandled tag template
While parsing a "template", corresponding to this C++ code:
Position:14:1
    template<typename T>
    ^

Warning: (SrcmlcppIgnoreElement) A cpp element of type "constructor" was stored as CppUnprocessed. Details follow
Warning: (SrcmlcppIgnoreElement) unhandled tag template
While parsing a "template", corresponding to this C++ code:
Position:16:1
    template<typename T>
    ^

Version & Platform

Mention

• the version of litgen you are using => 0.20.0
• the type of bindings you are using: => pybind11

Minimal reproducible example

cpp_code  = """
template <typename T>
class Hello {
    public:
        Hello();
        Hello(T c);

        T getC(void) {return this->c;}
    private:
        T c;

};

template<typename T>
inline Hello<T>::Hello(): c(0) {};
template<typename T>
inline Hello<T>::Hello(T c): c(c) {};
"""

import litgen
from litgen.demo import litgen_demo

options = litgen.LitgenOptions()
options.class_template_options.add_specialization(r"^Hello$", ["int", "float", "double"])
generated_code = litgen.generate_code(options, cpp_code)

---

Is there any possibility to not bind inline definitions or make the code work?
I have a lot of code like this.

[pthom]

You can either

• ignore these warnings (see doc on how to ignore warnings).
• split your headers in two parts : declaration in one file, then implementation in another separate header.
• put these implementation in a separate zone inside #ifdef (where the ifdef variable evaluates to true). Litgen by default ignores zones inside ifdef

[farinaleo]

Thanks for your help!! Ignoring the warnings seems to work but I'm facing another problem.

When a constructor of a "normal" class is defined with inline, the generated python looks strange. There is no warning or error but the python is not properly written and is not usable.

Minimal reproducible example

cpp_code  = """
class Hello {
    public:
        Hello();
        Hello(int c);

        int getC(void) {return this->c;}
    private:
        int c;

};

inline Hello::Hello(): c(0) {};
inline Hello::Hello(int c): c(c) {};
"""

import litgen
from litgen.demo import litgen_demo

options = litgen.LitgenOptions()
generated_code = litgen.generate_code(options, cpp_code)
print(generated_code.pydef_code)
print()
print(generated_code.stub_code)

Output

auto pyClassHello =
    py::class_<Hello>
        (m, "Hello", "")
    .def(py::init<>())
    .def(py::init<int>(),
        py::arg("c"))
    .def("get_c",
        &Hello::getC)
    ;


m.def("hello::_hello",
    py::overload_cast<>(Hello::Hello));

m.def("hello::_hello",
    py::overload_cast<int>(Hello::Hello), py::arg("c"));

class Hello:
    @overload
    def __init__(self) -> None:
        pass
    @overload
    def __init__(self, c: int) -> None:
        pass

    def get_c(self) -> int:
        pass

@overload
def hello::_hello() -> :
    pass
@overload
def hello::_hello(c: int) -> :
    pass

Is there any way to improve the output? Or do I have to use the #ifdef solution?

[pthom]

Is there any way to improve the output?

Yes, install from source.
With the source version I get:

binding C++ code

auto pyClassHello =
    py::class_<Hello>
        (m, "Hello", "")
    .def(py::init<>())
    .def(py::init<int>(),
        py::arg("c"))
    .def("get_c",
        &Hello::getC)
    ;

stub

class Hello:
    @overload
    def __init__(self) -> None:
        pass
    @overload
    def __init__(self, c: int) -> None:
        pass

    def get_c(self) -> int:
        pass

[farinaleo]

Thanks for answering!!

Your solution seems to be working great, but what do you mean by saying "install from source code?".

How have you got this result?

[pthom]

git clone https://github.com/pthom/litgen.git
cd litgen
pip install -v .