Add additional functions

[jurihock]

There are two colorbar functions in matplotlib: Figure.colorbar (as already implemented) and pyplot.colorbar (proposed in this pr). The mappable argument is mandatory in Figure.colorbar but optional in pyplot.colorbar, which internally defaults to pyplot.gci. See also matplotlib.pyplot.colorbar.

[soblin]

Hi @jurihock thanks for you PR !
Do you have example codes using these functions ? Or do you know any related examples in this page ?

https://matplotlib.org/stable/gallery/index.html

I want to add them in gallery/contrib folder for better coverage if possible.

[jurihock]

I plan to use it here instead of the current matplotlibcpp binding. But for that I need at least imshow and clim.

I'm not familiar with xtensor, but numcpp works fine to draw a 2D sombrero:

#include <matplotlibcpp17/pyplot.h>
#include <NumCpp.hpp>

namespace py = pybind11;

void test()
{
  py::scoped_interpreter guard{};

  const int size = 1000;
  const double sigma = 100;

  const auto i = nc::arange<double>(size) - (0.5 * size);
  const auto [x, y] = nc::meshgrid<double>(i, i);

  const auto xy = (nc::power(x, 2) + nc::power(y, 2)) / (-2.0 * nc::power(sigma, 2));

  const auto sombrero = nc::exp(xy) * (xy + 1.0) / (std::acos(-1.0) * nc::power(sigma, 4));
  const auto pysombrero = nc::pybindInterface::nc2pybind(sombrero);

  auto plot = matplotlibcpp17::pyplot::import();

  plot.imshow(Args(pysombrero), Kwargs(
    "extent"_a = py::make_tuple(-1, +1, -1, +1),
    "cmap"_a = "inferno"));

  plot.colorbar();

  plot.clim(py::make_tuple(-4e-10, +3e-9));

  plot.show();
}

Maybe something like that would be a suitable example for the gallery...

[soblin]

LGTM.

Then you can rename test() to main() and just push that code to gallery/contrib/sombrero,cpp without CMakeLists.txt.
Can you tell me the version or current commit hash of your numcpp ? Maybe I'll use numcpp instead of xtensor in the next release.
Also you can add that image to gallery/contrib/sombrero.png (maybe you need to add -f). Then I will review and merge this PR to latest main !

I think I'll add some examples from the official matplotlib gallery using imshow and other new functions later. Then I'll release next version.

[jurihock]

Okay, thanks!

I'm using the tagged NumCpp 2.8.0 version.

In order to build the provided example you may have to introduce following defines:

// disable boost
#define NUMCPP_NO_USE_BOOST

// enable nc::pybindInterface::nc2pybind
#define NUMCPP_INCLUDE_PYBIND_PYTHON_INTERFACE