Simpler handling of aggregate types

[h-2]

Say I have this type in C++:

struct Options
{
  bool opt1 = true;
  bool opt2 = false;
  bool opt3 = true;
  std::string opt4 = "fast";
  std::string opt5 = "default;
};

My current understanding is that to properly support this, I need to do:

nb::class_<Options>(m, "Options")
  .def_rw("opt1"     &Options::opt1)
  .def_rw("opt2"     &Options::opt2)
  .def_rw("opt3"     &Options::opt3)
  .def_rw("opt4"     &Options::opt4)
  .def_rw("opt5"     &Options::opt5)
  .def(nb::init<>())
  .def(nb::init<bool, bool, bool, std::string, std::string>());

Is that correct? Couldn't nanobind do all of this automatically for me if std::is_aggregate_v<> is true for the type?

[wjakob]

This would require some kind of reflection, which is not part of C++17.

Litgen can generate this kind of code by parsing header: https://pthom.github.io/litgen/litgen_book/00_00_intro.html

LLMs (Claude, ChatGPT) are also excellent at generating binding code these days, just copy-paste your header file and say "make nanobind bindings".