Homec4science
Diffusion akantu (features/50-using-gitpod-in-gitlab) Browse

rAKA/python/pybind11/py_aka_array.cc
df0ac98c3b81features/50-using…

Last Change
Raw File

py_aka_array.cc
View Options

/* -------------------------------------------------------------------------- */
#include "aka_array.hh"
/* -------------------------------------------------------------------------- */
#include <pybind11/numpy.h>
#include <pybind11/pybind11.h>
/* -------------------------------------------------------------------------- */
namespace py = pybind11;
namespace _aka = akantu;
namespace akantu {
template <typename T> class ArrayProxy : public Array<T> {
protected:
// deallocate the memory
void deallocate() override final {}
// allocate the memory
void allocate(__attribute__((unused)) UInt size,
__attribute__((unused)) UInt nb_component) override final {}
// allocate and initialize the memory
void allocate(__attribute__((unused)) UInt size,
__attribute__((unused)) UInt nb_component,
__attribute__((unused)) const T & value) override final {}
public:
ArrayProxy(T * data, UInt size, UInt nb_component) {
this->values = data;
this->size_ = size;
this->nb_component = nb_component;
}
ArrayProxy(const Array<T> & src) {
this->values = src.storage();
this->size_ = src.size();
this->nb_component = src.getNbComponent();
}
void resize(__attribute__((unused)) UInt size,
__attribute__((unused)) const T & val) override final {
AKANTU_EXCEPTION("cannot resize a temporary array");
}
void resize(__attribute__((unused)) UInt new_size) override final {
AKANTU_EXCEPTION("cannot resize a temporary array");
}
void reserve(__attribute__((unused)) UInt new_size) override final {
AKANTU_EXCEPTION("cannot resize a temporary array");
}
};
} // namespace akantu
namespace pybind11 {
namespace detail {
/* ------------------------------------------------------------------------ */
template <typename T>
py::handle aka_array_cast(const _aka::Array<T> & src,
py::handle base = handle(), bool writeable = true) {
constexpr ssize_t elem_size = sizeof(T);
array a;
a = array({src.size(), src.getNbComponent()},
{elem_size * src.getNbComponent(), elem_size}, src.storage(),
base);
if (not writeable)
array_proxy(a.ptr())->flags &= ~detail::npy_api::NPY_ARRAY_WRITEABLE_;
return a.release();
}
template <typename T> struct type_caster<_aka::Array<T>> {
protected:
using py_array = array_t<T, array::forcecast | array::c_style>;
using type = _aka::Array<T>;
public:
PYBIND11_TYPE_CASTER(type, _("Array<T>"));
// static constexpr auto name =
// _("numpy.ndarray[") + npy_format_descriptor<T>::name +
// _(", flags.writeable") + _(", flags.c_contiguous") + _("]");
/**
* Conversion part 1 (Python->C++)
*/
bool load(handle src, bool convert) {
bool need_copy = not isinstance<py_array>(src);
auto && fits = [&](auto && a) {
auto && dims = copy_or_ref.ndim();
if (dims < 1 || dims > 2)
return false;
return true;
};
if (not need_copy) {
// We don't need a converting copy, but we also need to check whether
// the strides are compatible with the Ref's stride requirements
py_array aref = reinterpret_borrow<py_array>(src);
if (not fits(aref)) {
return false;
}
copy_or_ref = std::move(aref);
} else {
if (not convert) {
return false;
}
auto copy = py_array::ensure(src);
if (not copy) {
return false;
}
if (not fits(copy)) {
return false;
}
copy_or_ref = std::move(py_array::ensure(src));
loader_life_support::add_patient(copy_or_ref);
}
array_proxy.reset(new _aka::ArrayProxy<T>> (copy_or_ref.mutable_data(),
copy_or_ref.shape(0),
copy_or_ref.shape(1)));
return true;
}
// operator _aka::Array<T> *() { return array_proxy.get(); }
// operator _aka::Array<T> &() { return *array_proxy; }
// template <typename _T>
// using cast_op_type = pybind11::detail::cast_op_type<_T>;
/**
* Conversion part 2 (C++ -> Python)
*/
static handle cast(const _aka::Array<T> & src, return_value_policy policy,
handle parent) {
switch (policy) {
case return_value_policy::copy:
return aka_array_cast<T>(src);
case return_value_policy::reference_internal:
return aka_array_cast<T>(src, parent);
case return_value_policy::reference:
case return_value_policy::automatic:
case return_value_policy::automatic_reference:
return aka_array_cast<T>(src, none());
default:
pybind11_fail("Invalid return_value_policy for ArrayProxy type");
}
}
protected:
std::unique_ptr<_aka::Array<T>> array_proxy;
py_array copy_or_ref;
};
/**
* Type caster for Vector classes
*/
template <template <typename> class V, typename T> struct type_caster<V<T>> {
using type = V<T>;
template <typename U>
using array_type = array_t<U, array::c_style | array::forcecast>;
public:
PYBIND11_TYPE_CASTER(type, _("Vector<T>"));
/**
* Conversion part 1 (Python->C++): convert a PyObject into a Vector
* instance or return false upon failure. The second argument
* indicates whether implicit conversions should be applied.
*/
bool load(handle src, bool convert) {
if (!convert && !array_type<typename type::value_type>::check_(src))
return false;
auto buf = array_type<typename type::value_type>::ensure(src);
value.move(_aka::VectorProxy<T>(buf));
return true;
}
/**
* Conversion part 2 (C++ -> Python): convert a grid instance into
* a Python object. The second and third arguments are used to
* indicate the return value policy and parent object (for
* ``return_value_policy::reference_internal``) and are generally
* ignored by implicit casters.
*
* TODO: do not ignore policy (see pybind11/eigen.h)
*/
static handle cast(const type & src, return_value_policy /* policy */,
handle /*parent*/) {
// none() passed as parent to get a correct nocopy
auto a = array_type<typename type::value_type>(src.size(), src.storage(),
none());
return a.release();
}
};
} // namespace detail
} // namespace pybind11
namespace {
py::module & register_arrays(py::module & mod) {
// py::class_<_aka::Vector<bool>>(mod, "VectorBool");
// py::class_<_aka::Vector<double>>(mod, "VectorReal");
return mod;
}
} // namespace
c4science · Help