ic
/
magnum-bindings
6
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
123 Commits
4 Branches
2 Tags
2.4 MiB
Tree: cfce76dde4
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from 'cfce76dde4'
${ noResults }
magnum-bindings/src/python/magnum/magnum.cpp

293 lines
12 KiB
Raw Normal View History Unescape

Initial code for the Python bindings, adding math angle types. Only the double ones, exposed as floats, because the extra ALU required by doubles is negligible to function call overhead. It'll be different for non-scalar types, but here I use this.
7 years ago
/*
This file is part of Magnum.
Copyright © 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017, 2018, 2019
Vladimír Vondruš <mosra@centrum.cz>
Permission is hereby granted, free of charge, to any person obtaining a
copy of this software and associated documentation files (the "Software"),
to deal in the Software without restriction, including without limitation
the rights to use, copy, modify, merge, publish, distribute, sublicense,
and/or sell copies of the Software, and to permit persons to whom the
Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included
in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
DEALINGS IN THE SOFTWARE.
*/
python: bind PixelFormat and PixelStorage. Just the uncompressed ones for now.
7 years ago
#include <pybind11/operators.h>
Initial code for the Python bindings, adding math angle types. Only the double ones, exposed as floats, because the extra ALU required by doubles is negligible to function call overhead. It'll be different for non-scalar types, but here I use this.
7 years ago
#include <pybind11/pybind11.h>
python: store memory owner references in the holder type instead. Compared to having to subclass every type that can reference external data, this has several advantages for 3rd party binding code: * it doesn't need to worry about the additional type when binding function arguments (currently it had to provide lambdas that accept the PyFoo subtype instead of just Foo) * and it can now easily bind those types also for function return values and properties -- the return type doesn't need to be subclassed (which in case of move-only types is practically impossible) but instead just wrapped in a holder along with the memory owner object reference The new holders also assert that memory owner is always specified unless the data is empty.
7 years ago
#include <Corrade/Containers/StridedArrayView.h>
#include <Magnum/ImageView.h>
python: add generic MeshPrimitive and MeshIndexType enums.
7 years ago
#include <Magnum/Mesh.h>
python: bind PixelFormat and PixelStorage. Just the uncompressed ones for now.
7 years ago
#include <Magnum/PixelFormat.h>
#include <Magnum/PixelStorage.h>
Initial code for the Python bindings, adding math angle types. Only the double ones, exposed as floats, because the extra ALU required by doubles is negligible to function call overhead. It'll be different for non-scalar types, but here I use this.
7 years ago
python: store memory owner references in the holder type instead. Compared to having to subclass every type that can reference external data, this has several advantages for 3rd party binding code: * it doesn't need to worry about the additional type when binding function arguments (currently it had to provide lambdas that accept the PyFoo subtype instead of just Foo) * and it can now easily bind those types also for function return values and properties -- the return type doesn't need to be subclassed (which in case of move-only types is practically impossible) but instead just wrapped in a holder along with the memory owner object reference The new holders also assert that memory owner is always specified unless the data is empty.
7 years ago
#include "Corrade/Python.h"
#include "Corrade/Containers/Python.h"
python: expose (Mutable)ImageView.
7 years ago
#include "Magnum/Python.h"
Initial code for the Python bindings, adding math angle types. Only the double ones, exposed as floats, because the extra ALU required by doubles is negligible to function call overhead. It'll be different for non-scalar types, but here I use this.
7 years ago
#include "magnum/bootstrap.h"
python: put everything into a single module if magnum is built statically. That makes it much easier to handle in that case (because that's the whole point of static builds) and also nicely circumvents any issues with duplicated global data when a static lib would be linked to multiple dynamic libraries.
7 years ago
#ifdef MAGNUM_BUILD_STATIC
#include "magnum/staticconfigure.h"
#endif
Initial code for the Python bindings, adding math angle types. Only the double ones, exposed as floats, because the extra ALU required by doubles is negligible to function call overhead. It'll be different for non-scalar types, but here I use this.
7 years ago
namespace py = pybind11;
python: add generic MeshPrimitive and MeshIndexType enums.
7 years ago
namespace magnum { namespace {
python: expose (Mutable)ImageView.
7 years ago
template<UnsignedInt dimensions, class T> struct PyDimensionTraits;
template<class T> struct PyDimensionTraits<1, T> {
typedef T VectorType;
static VectorType from(const Math::Vector<1, T>& vec) { return vec[0]; }
};
template<class T> struct PyDimensionTraits<2, T> {
typedef Math::Vector2<T> VectorType;
static VectorType from(const Math::Vector<2, T>& vec) { return vec; }
};
template<class T> struct PyDimensionTraits<3, T> {
typedef Math::Vector3<T> VectorType;
static VectorType from(const Math::Vector<3, T>& vec) { return vec; }
};
python: store memory owner references in the holder type instead. Compared to having to subclass every type that can reference external data, this has several advantages for 3rd party binding code: * it doesn't need to worry about the additional type when binding function arguments (currently it had to provide lambdas that accept the PyFoo subtype instead of just Foo) * and it can now easily bind those types also for function return values and properties -- the return type doesn't need to be subclassed (which in case of move-only types is practically impossible) but instead just wrapped in a holder along with the memory owner object reference The new holders also assert that memory owner is always specified unless the data is empty.
7 years ago
template<class T> void imageView(py::class_<T, PyImageViewHolder<T>>& c) {
python: expose (Mutable)ImageView.
7 years ago
/*
Missing APIs:
Type, ErasedType, Dimensions
*/
c
/* Constructors */
python: store memory owner references in the holder type instead. Compared to having to subclass every type that can reference external data, this has several advantages for 3rd party binding code: * it doesn't need to worry about the additional type when binding function arguments (currently it had to provide lambdas that accept the PyFoo subtype instead of just Foo) * and it can now easily bind those types also for function return values and properties -- the return type doesn't need to be subclassed (which in case of move-only types is practically impossible) but instead just wrapped in a holder along with the memory owner object reference The new holders also assert that memory owner is always specified unless the data is empty.
7 years ago
.def(py::init([](const PixelStorage& storage, PixelFormat format, const typename PyDimensionTraits<T::Dimensions, Int>::VectorType& size, const Containers::ArrayView<typename T::Type>& data) {
return pyImageViewHolder(T{storage, format, size, data}, pyObjectHolderFor<Containers::PyArrayViewHolder>(data).owner);
python: expose (Mutable)ImageView.
7 years ago
}), "Constructor")
python: store memory owner references in the holder type instead. Compared to having to subclass every type that can reference external data, this has several advantages for 3rd party binding code: * it doesn't need to worry about the additional type when binding function arguments (currently it had to provide lambdas that accept the PyFoo subtype instead of just Foo) * and it can now easily bind those types also for function return values and properties -- the return type doesn't need to be subclassed (which in case of move-only types is practically impossible) but instead just wrapped in a holder along with the memory owner object reference The new holders also assert that memory owner is always specified unless the data is empty.
7 years ago
.def(py::init([](PixelFormat format, const typename PyDimensionTraits<T::Dimensions, Int>::VectorType& size, const Containers::ArrayView<typename T::Type>& data) {
return pyImageViewHolder(T{format, size, data}, pyObjectHolderFor<Containers::PyArrayViewHolder>(data).owner);
python: expose (Mutable)ImageView.
7 years ago
}), "Constructor")
.def(py::init([](const PixelStorage& storage, PixelFormat format, const typename PyDimensionTraits<T::Dimensions, Int>::VectorType& size) {
python: store memory owner references in the holder type instead. Compared to having to subclass every type that can reference external data, this has several advantages for 3rd party binding code: * it doesn't need to worry about the additional type when binding function arguments (currently it had to provide lambdas that accept the PyFoo subtype instead of just Foo) * and it can now easily bind those types also for function return values and properties -- the return type doesn't need to be subclassed (which in case of move-only types is practically impossible) but instead just wrapped in a holder along with the memory owner object reference The new holders also assert that memory owner is always specified unless the data is empty.
7 years ago
return T{storage, format, size};
python: expose (Mutable)ImageView.
7 years ago
}), "Construct an empty view")
.def(py::init([](PixelFormat format, const typename PyDimensionTraits<T::Dimensions, Int>::VectorType& size) {
python: store memory owner references in the holder type instead. Compared to having to subclass every type that can reference external data, this has several advantages for 3rd party binding code: * it doesn't need to worry about the additional type when binding function arguments (currently it had to provide lambdas that accept the PyFoo subtype instead of just Foo) * and it can now easily bind those types also for function return values and properties -- the return type doesn't need to be subclassed (which in case of move-only types is practically impossible) but instead just wrapped in a holder along with the memory owner object reference The new holders also assert that memory owner is always specified unless the data is empty.
7 years ago
return T{format, size};
python: expose (Mutable)ImageView.
7 years ago
}), "Construct an empty view")
/* Properties */
.def_property_readonly("storage", &T::storage, "Storage of pixel data")
.def_property_readonly("format", &T::format, "Format of pixel data")
.def_property_readonly("pixel_size", &T::pixelSize, "Pixel size (in bytes)")
.def_property_readonly("size", [](T& self) {
return PyDimensionTraits<T::Dimensions, Int>::from(self.size());
}, "Image size")
.def_property("data", [](T& self) {
python: store memory owner references in the holder type instead. Compared to having to subclass every type that can reference external data, this has several advantages for 3rd party binding code: * it doesn't need to worry about the additional type when binding function arguments (currently it had to provide lambdas that accept the PyFoo subtype instead of just Foo) * and it can now easily bind those types also for function return values and properties -- the return type doesn't need to be subclassed (which in case of move-only types is practically impossible) but instead just wrapped in a holder along with the memory owner object reference The new holders also assert that memory owner is always specified unless the data is empty.
7 years ago
return Containers::pyArrayViewHolder(self.data(), pyObjectHolderFor<PyImageViewHolder>(self).owner);
}, [](T& self, const Containers::ArrayView<typename T::Type>& data) {
python: expose (Mutable)ImageView.
7 years ago
self.setData(data);
python: store memory owner references in the holder type instead. Compared to having to subclass every type that can reference external data, this has several advantages for 3rd party binding code: * it doesn't need to worry about the additional type when binding function arguments (currently it had to provide lambdas that accept the PyFoo subtype instead of just Foo) * and it can now easily bind those types also for function return values and properties -- the return type doesn't need to be subclassed (which in case of move-only types is practically impossible) but instead just wrapped in a holder along with the memory owner object reference The new holders also assert that memory owner is always specified unless the data is empty.
7 years ago
pyObjectHolderFor<PyImageViewHolder>(self).owner =
pyObjectHolderFor<Containers::PyArrayViewHolder>(data).owner;
python: expose (Mutable)ImageView.
7 years ago
}, "Image data")
.def_property_readonly("pixels", [](T& self) {
python: store memory owner references in the holder type instead. Compared to having to subclass every type that can reference external data, this has several advantages for 3rd party binding code: * it doesn't need to worry about the additional type when binding function arguments (currently it had to provide lambdas that accept the PyFoo subtype instead of just Foo) * and it can now easily bind those types also for function return values and properties -- the return type doesn't need to be subclassed (which in case of move-only types is practically impossible) but instead just wrapped in a holder along with the memory owner object reference The new holders also assert that memory owner is always specified unless the data is empty.
7 years ago
return Containers::pyArrayViewHolder(self.pixels(), pyObjectHolderFor<PyImageViewHolder>(self).owner);
python: expose (Mutable)ImageView.
7 years ago
}, "View on pixel data")
python: store memory owner references in the holder type instead. Compared to having to subclass every type that can reference external data, this has several advantages for 3rd party binding code: * it doesn't need to worry about the additional type when binding function arguments (currently it had to provide lambdas that accept the PyFoo subtype instead of just Foo) * and it can now easily bind those types also for function return values and properties -- the return type doesn't need to be subclassed (which in case of move-only types is practically impossible) but instead just wrapped in a holder along with the memory owner object reference The new holders also assert that memory owner is always specified unless the data is empty.
7 years ago
.def_property_readonly("owner", [](T& self) {
return pyObjectHolderFor<PyImageViewHolder>(self).owner;
}, "Memory owner");
python: expose (Mutable)ImageView.
7 years ago
}
python: store memory owner references in the holder type instead. Compared to having to subclass every type that can reference external data, this has several advantages for 3rd party binding code: * it doesn't need to worry about the additional type when binding function arguments (currently it had to provide lambdas that accept the PyFoo subtype instead of just Foo) * and it can now easily bind those types also for function return values and properties -- the return type doesn't need to be subclassed (which in case of move-only types is practically impossible) but instead just wrapped in a holder along with the memory owner object reference The new holders also assert that memory owner is always specified unless the data is empty.
7 years ago
template<class T> void imageViewFromMutable(py::class_<T, PyImageViewHolder<T>>& c) {
python: expose (Mutable)ImageView.
7 years ago
c
python: store memory owner references in the holder type instead. Compared to having to subclass every type that can reference external data, this has several advantages for 3rd party binding code: * it doesn't need to worry about the additional type when binding function arguments (currently it had to provide lambdas that accept the PyFoo subtype instead of just Foo) * and it can now easily bind those types also for function return values and properties -- the return type doesn't need to be subclassed (which in case of move-only types is practically impossible) but instead just wrapped in a holder along with the memory owner object reference The new holders also assert that memory owner is always specified unless the data is empty.
7 years ago
.def(py::init([](const BasicMutableImageView<T::Dimensions>& other) {
return pyImageViewHolder(BasicImageView<T::Dimensions>(other), pyObjectHolderFor<PyImageViewHolder>(other).owner);
python: expose (Mutable)ImageView.
7 years ago
}), "Constructor");
}
python: add generic MeshPrimitive and MeshIndexType enums.
7 years ago
void magnum(py::module& m) {
py::enum_<MeshPrimitive>{m, "MeshPrimitive", "Mesh primitive type"}
.value("POINTS", MeshPrimitive::Points)
.value("LINES", MeshPrimitive::Lines)
.value("LINE_LOOP", MeshPrimitive::LineLoop)
.value("LINE_STRIP", MeshPrimitive::LineStrip)
.value("TRIANGLES", MeshPrimitive::Triangles)
.value("TRIANGLE_STRIP", MeshPrimitive::TriangleStrip)
.value("TRIANGLE_FAN", MeshPrimitive::TriangleFan);
py::enum_<MeshIndexType>{m, "MeshIndexType", "Mesh index type"}
.value("UNSIGNED_BYTE", MeshIndexType::UnsignedByte)
.value("UNSIGNED_SHORT", MeshIndexType::UnsignedShort)
.value("UNSIGNED_INT", MeshIndexType::UnsignedInt);
python: bind PixelFormat and PixelStorage. Just the uncompressed ones for now.
7 years ago
py::enum_<PixelFormat>{m, "PixelFormat", "Format of pixel data"}
.value("R8UNORM", PixelFormat::R8Unorm)
.value("RG8UNORM", PixelFormat::RG8Unorm)
.value("RGB8UNORM", PixelFormat::RGB8Unorm)
.value("RGBA8UNORM", PixelFormat::RGBA8Unorm)
.value("R8SNORM", PixelFormat::R8Snorm)
.value("RG8SNORM", PixelFormat::RG8Snorm)
.value("RGB8SNORM", PixelFormat::RGB8Snorm)
.value("RGBA8SNORM", PixelFormat::RGBA8Snorm)
.value("R8UI", PixelFormat::R8UI)
.value("RG8UI", PixelFormat::RG8UI)
.value("RGB8UI", PixelFormat::RGB8UI)
.value("RGBA8UI", PixelFormat::RGBA8UI)
.value("R8I", PixelFormat::R8I)
.value("RG8I", PixelFormat::RG8I)
.value("RGB8I", PixelFormat::RGB8I)
.value("RGBA8I", PixelFormat::RGBA8I)
.value("R16UNORM", PixelFormat::R16Unorm)
.value("RG16UNORM", PixelFormat::RG16Unorm)
.value("RGB16UNORM", PixelFormat::RGB16Unorm)
.value("RGBA16UNORM", PixelFormat::RGBA16Unorm)
.value("R16SNORM", PixelFormat::R16Snorm)
.value("RG16SNORM", PixelFormat::RG16Snorm)
.value("RGB16SNORM", PixelFormat::RGB16Snorm)
.value("RGBA16SNORM", PixelFormat::RGBA16Snorm)
.value("R16UI", PixelFormat::R16UI)
.value("RG16UI", PixelFormat::RG16UI)
.value("RGB16UI", PixelFormat::RGB16UI)
.value("RGBA16UI", PixelFormat::RGBA16UI)
.value("R16I", PixelFormat::R16I)
.value("RG16I", PixelFormat::RG16I)
.value("RGB16I", PixelFormat::RGB16I)
.value("RGBA16I", PixelFormat::RGBA16I)
.value("R32UI", PixelFormat::R32UI)
.value("RG32UI", PixelFormat::RG32UI)
.value("RGB32UI", PixelFormat::RGB32UI)
.value("RGBA32UI", PixelFormat::RGBA32UI)
.value("R32I", PixelFormat::R32I)
.value("RG32I", PixelFormat::RG32I)
.value("RGB32I", PixelFormat::RGB32I)
.value("RGBA32I", PixelFormat::RGBA32I)
.value("R16F", PixelFormat::R16F)
.value("RG16F", PixelFormat::RG16F)
.value("RGB16F", PixelFormat::RGB16F)
.value("RGBA16F", PixelFormat::RGBA16F)
.value("R32F", PixelFormat::R32F)
.value("RG32F", PixelFormat::RG32F)
.value("RGB32F", PixelFormat::RGB32F)
.value("RGBA32F", PixelFormat::RGBA32F);
py::class_<PixelStorage>{m, "PixelStorage", "Pixel storage parameters"}
.def(py::init(), "Default constructor")
/* Comparison */
.def(py::self == py::self, "Equality comparison")
.def(py::self != py::self, "Non-equality comparison")
/* Properties */
.def_property("alignment",
&PixelStorage::alignment, &PixelStorage::setAlignment, "Row alignment")
.def_property("row_length",
&PixelStorage::rowLength, &PixelStorage::setRowLength, "Row length")
.def_property("image_height",
&PixelStorage::imageHeight, &PixelStorage::setImageHeight, "Image height")
.def_property("skip",
&PixelStorage::skip, &PixelStorage::setSkip, "Pixel, row and image skip");
python: expose (Mutable)ImageView.
7 years ago
python: store memory owner references in the holder type instead. Compared to having to subclass every type that can reference external data, this has several advantages for 3rd party binding code: * it doesn't need to worry about the additional type when binding function arguments (currently it had to provide lambdas that accept the PyFoo subtype instead of just Foo) * and it can now easily bind those types also for function return values and properties -- the return type doesn't need to be subclassed (which in case of move-only types is practically impossible) but instead just wrapped in a holder along with the memory owner object reference The new holders also assert that memory owner is always specified unless the data is empty.
7 years ago
py::class_<ImageView1D, PyImageViewHolder<ImageView1D>> imageView1D{m, "ImageView1D", "One-dimensional image view"};
py::class_<ImageView2D, PyImageViewHolder<ImageView2D>> imageView2D{m, "ImageView2D", "Two-dimensional image view"};
py::class_<ImageView3D, PyImageViewHolder<ImageView3D>> imageView3D{m, "ImageView3D", "Three-dimensional image view"};
py::class_<MutableImageView1D, PyImageViewHolder<MutableImageView1D>> mutableImageView1D{m, "MutableImageView1D", "One-dimensional mutable image view"};
py::class_<MutableImageView2D, PyImageViewHolder<MutableImageView2D>> mutableImageView2D{m, "MutableImageView2D", "Two-dimensional mutable image view"};
py::class_<MutableImageView3D, PyImageViewHolder<MutableImageView3D>> mutableImageView3D{m, "MutableImageView3D", "Three-dimensional mutable image view"};
python: expose (Mutable)ImageView.
7 years ago
imageView(imageView1D);
imageView(imageView2D);
imageView(imageView3D);
imageView(mutableImageView1D);
imageView(mutableImageView2D);
imageView(mutableImageView3D);
imageViewFromMutable(imageView1D);
imageViewFromMutable(imageView2D);
imageViewFromMutable(imageView3D);
python: add generic MeshPrimitive and MeshIndexType enums.
7 years ago
}
}}
Initial code for the Python bindings, adding math angle types. Only the double ones, exposed as floats, because the extra ALU required by doubles is negligible to function call overhead. It'll be different for non-scalar types, but here I use this.
7 years ago
PYBIND11_MODULE(_magnum, m) {
python: expose (Mutable)ImageView.
7 years ago
/* We need ArrayView for images */
py::module::import("corrade.containers");
Initial code for the Python bindings, adding math angle types. Only the double ones, exposed as floats, because the extra ALU required by doubles is negligible to function call overhead. It'll be different for non-scalar types, but here I use this.
7 years ago
m.doc() = "Root Magnum module";
py::module math = m.def_submodule("math");
magnum::math(m, math);
python: put everything into a single module if magnum is built statically. That makes it much easier to handle in that case (because that's the whole point of static builds) and also nicely circumvents any issues with duplicated global data when a static lib would be linked to multiple dynamic libraries.
7 years ago
python: bind PixelFormat and PixelStorage. Just the uncompressed ones for now.
7 years ago
/* These need stuff from math, so need to be called after */
magnum::magnum(m);
python: put everything into a single module if magnum is built statically. That makes it much easier to handle in that case (because that's the whole point of static builds) and also nicely circumvents any issues with duplicated global data when a static lib would be linked to multiple dynamic libraries.
7 years ago
/* In case Magnum is a bunch of static libraries, put everything into a
single shared lib to make it easier to install (which is the point of
static builds) and avoid issues with multiply-defined global symbols.
These need to be defined in the order they depend on. */
#ifdef MAGNUM_BUILD_STATIC
#ifdef Magnum_GL_FOUND
py::module gl = m.def_submodule("gl");
magnum::gl(gl);
#endif
#ifdef Magnum_SceneGraph_FOUND
py::module scenegraph = m.def_submodule("scenegraph");
magnum::scenegraph(scenegraph);
#endif
#ifdef Magnum_Trade_FOUND
py::module trade = m.def_submodule("trade");
magnum::trade(trade);
#endif
#ifdef Magnum_MeshTools_FOUND
/* Depends on trade and gl */
py::module meshtools = m.def_submodule("meshtools");
magnum::meshtools(meshtools);
#endif
#ifdef Magnum_Primitives_FOUND
/* Depends on trade */
py::module primitives = m.def_submodule("primitives");
magnum::primitives(primitives);
#endif
#ifdef Magnum_Shaders_FOUND
/* Depends on gl */
py::module shaders = m.def_submodule("shaders");
magnum::shaders(shaders);
#endif
/* Keep the doc in sync with platform/__init__.py */
py::module platform = m.def_submodule("platform");
platform.doc() = "Platform-specific application and context creation";
#ifdef Magnum_GlfwApplication_FOUND
py::module glfw = platform.def_submodule("glfw");
magnum::platform::glfw(glfw);
#endif
#ifdef Magnum_Sdl2Application_FOUND
py::module sdl2 = platform.def_submodule("sdl2");
magnum::platform::sdl2(sdl2);
#endif
#ifdef Magnum_WindowlessEglApplication_FOUND
py::module egl = platform.def_submodule("egl");
magnum::platform::egl(egl);
#endif
#ifdef Magnum_WindowlessGlxApplication_FOUND
py::module glx = platform.def_submodule("glx");
magnum::platform::glx(glx);
#endif
#endif
Initial code for the Python bindings, adding math angle types. Only the double ones, exposed as floats, because the extra ALU required by doubles is negligible to function call overhead. It'll be different for non-scalar types, but here I use this.
7 years ago
}