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.
385 Commits
4 Branches
2 Tags
2.4 MiB
Tree: 66bb4451df
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from '66bb4451df'
${ noResults }
magnum-bindings/src/python/magnum/magnum.cpp

423 lines
18 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.
Updated copyright year.
6 years ago
Copyright © 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017, 2018, 2019,
Updated copyright year.
4 years ago
2020, 2021, 2022 Vladimír Vondruš <mosra@centrum.cz>
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
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>
python: expose Image. Owning counterpart to ImageView.
7 years ago
#include <Magnum/Image.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 <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>
python: expose textures and related APIs.
7 years ago
#include <Magnum/Sampler.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: rename all Python.h headers to PythonBindings.h. It conveys clearer what those headers are for and avoids issues where shitty IDE indexers (Eclipse?) confuse this with Python's Python.h.
6 years ago
#include "Corrade/PythonBindings.h"
#include "Corrade/Containers/PythonBindings.h"
python: basic support for arbitrary types in strided array views. Not ArrayView yet, and also no documentation on this whatsoever. That'll come next. Also not everything works with arbitrary types yet, converting from buffer protocol doesn't remember the format and conversion to bytes doesn't take the actual type size into account either.
5 years ago
#include "Corrade/Containers/StridedArrayViewPythonBindings.h"
python: rename all Python.h headers to PythonBindings.h. It conveys clearer what those headers are for and avoids issues where shitty IDE indexers (Eclipse?) confuse this with Python's Python.h.
6 years ago
#include "Magnum/PythonBindings.h"
python: expose (Mutable)ImageView.
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
#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 Image. Owning counterpart to ImageView.
7 years ago
template<class T> void image(py::class_<T>& c) {
c
/* Constructors. Only the ones taking the generic format and *not*
taking an Array, as Python has no way to "move" it in */
.def(py::init<const PixelStorage&, PixelFormat>(), "Construct an image placeholder")
.def(py::init<PixelFormat>(), "Construct an image placeholder")
/* Properties */
.def_property_readonly("storage", &T::storage, "Storage of pixel data")
.def_property_readonly("format", &T::format, "Format of pixel data")
/** @todo formatExtra() */
.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_readonly("data", [](T& self) {
return Containers::pyArrayViewHolder(self.data(), self.data() ? py::cast(self) : py::none{});
}, "Image data")
.def_property_readonly("pixels", [](T& self) {
python: basic support for arbitrary types in strided array views. Not ArrayView yet, and also no documentation on this whatsoever. That'll come next. Also not everything works with arbitrary types yet, converting from buffer protocol doesn't remember the format and conversion to bytes doesn't take the actual type size into account either.
5 years ago
return Containers::pyArrayViewHolder(Containers::PyStridedArrayView<T::Dimensions + 1, const char>{self.pixels()}, self.data() ? py::cast(self) : py::none{});
python: expose Image. Owning counterpart to 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
template<class T> void imageView(py::class_<T, PyImageViewHolder<T>>& c) {
python: expose (Mutable)ImageView.
7 years ago
/*
Missing APIs:
Type, ErasedType, Dimensions
*/
python: expose Image. Owning counterpart to ImageView.
7 years ago
py::implicitly_convertible<Image<T::Dimensions>, T>();
python: expose (Mutable)ImageView.
7 years ago
c
python: sad workaround for a nasty problem deep inside pybind. :(
7 years ago
/* Constructors. The variants *not* taking an array view have to be
first, otherwise things fail on systems that don't have numpy
installed:
===================================================================
ERROR: test_init_empty (test.test.ImageView)
-------------------------------------------------------------------
Traceback (most recent call last):
File ".../magnum/test/test.py", line 102, in test_init_empty
b = ImageView2D(storage, PixelFormat.R32F, (8, 8))
ModuleNotFoundError: No module named 'numpy'
This is because of the order in which pybind processes arguments ---
it would first try to match the (PixelFormat, Vector2i, ArrayView)
variant and *somehow* getting all the way to the third argument,
where, because ArrayView is marked as implicitly convertible from
py::array for numpy compatibility, it ends up doing this in numpy.h:
module m = module::import("numpy.core.multiarray");
auto c = m.attr("_ARRAY_API");
Wonderful, isn't it. */
python: expose (Mutable)ImageView.
7 years ago
.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")
python: sad workaround for a nasty problem deep inside pybind. :(
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);
}), "Constructor")
.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);
}), "Constructor")
python: expose Image. Owning counterpart to ImageView.
7 years ago
.def(py::init([](Image<T::Dimensions>& image) {
return pyImageViewHolder(T{image}, image.data() ? py::cast(image) : py::none{});
}), "Construct a view on an image")
python: expose basic image import in trade. Quite a huge commit because it all needs to be tested together.
7 years ago
.def(py::init([](const ImageView<T::Dimensions, typename T::Type>& other) {
return pyImageViewHolder(ImageView<T::Dimensions, typename T::Type>(other), pyObjectHolderFor<PyImageViewHolder>(other).owner);
}), "Construct from any type convertible to an image view")
python: expose (Mutable)ImageView.
7 years ago
/* 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: basic support for arbitrary types in strided array views. Not ArrayView yet, and also no documentation on this whatsoever. That'll come next. Also not everything works with arbitrary types yet, converting from buffer protocol doesn't remember the format and conversion to bytes doesn't take the actual type size into account either.
5 years ago
return Containers::pyArrayViewHolder(Containers::PyStridedArrayView<T::Dimensions + 1, typename T::Type>{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: make conversion from a mutable image view implicit.
7 years ago
py::implicitly_convertible<BasicMutableImageView<T::Dimensions>, T>();
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 basic image import in trade. Quite a huge commit because it all needs to be tested together.
7 years ago
}), "Construct from a mutable view");
python: expose (Mutable)ImageView.
7 years ago
}
python: make this compatible with pybind11 2.6. Pybind changed py::module to py::module_ in order to support C++ modules, adapting to that change. It still builds on older versions, but we're using only the new APIs.
6 years ago
void magnum(py::module_& m) {
python: expose corrade and magnum compilation defines.
7 years ago
m.attr("BUILD_STATIC") =
#ifdef MAGNUM_BUILD_STATIC
true
#else
false
#endif
;
m.attr("TARGET_GL") =
#ifdef MAGNUM_TARGET_GL
true
#else
false
#endif
;
m.attr("TARGET_GLES") =
#ifdef MAGNUM_TARGET_GLES
true
#else
false
#endif
;
m.attr("TARGET_GLES2") =
#ifdef MAGNUM_TARGET_GLES2
true
#else
false
#endif
;
/** @todo do we need TARGET_GLES3? i hope not */
m.attr("TARGET_WEBGL") =
#ifdef MAGNUM_TARGET_WEBGL
true
#else
false
#endif
;
/* TARGET_DESKTOP_GLES, TARGET_HEADLESS skipped as they make sense only
on native side (affecting what the builtin utilities use), not really in
Python */
m.attr("TARGET_VK") =
#ifdef MAGNUM_TARGET_VK
true
#else
false
#endif
;
python: add generic MeshPrimitive and MeshIndexType enums.
7 years ago
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"}
python: use an underscore for the less readable PixelFormat values.
7 years ago
.value("R8_UNORM", PixelFormat::R8Unorm)
.value("RG8_UNORM", PixelFormat::RG8Unorm)
.value("RGB8_UNORM", PixelFormat::RGB8Unorm)
.value("RGBA8_UNORM", PixelFormat::RGBA8Unorm)
.value("R8_SNORM", PixelFormat::R8Snorm)
.value("RG8_SNORM", PixelFormat::RG8Snorm)
.value("RGB8_SNORM", PixelFormat::RGB8Snorm)
.value("RGBA8_SNORM", PixelFormat::RGBA8Snorm)
.value("R8_SRGB", PixelFormat::R8Srgb)
.value("RG8_SRGB", PixelFormat::RG8Srgb)
.value("RGB8_SRGB", PixelFormat::RGB8Srgb)
.value("RGBA8_SRGB", PixelFormat::RGBA8Srgb)
python: bind PixelFormat and PixelStorage. Just the uncompressed ones for now.
7 years ago
.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)
python: use an underscore for the less readable PixelFormat values.
7 years ago
.value("R16_UNORM", PixelFormat::R16Unorm)
.value("RG16_UNORM", PixelFormat::RG16Unorm)
.value("RGB16_UNORM", PixelFormat::RGB16Unorm)
.value("RGBA16_UNORM", PixelFormat::RGBA16Unorm)
.value("R16_SNORM", PixelFormat::R16Snorm)
.value("RG16_SNORM", PixelFormat::RG16Snorm)
.value("RGB16_SNORM", PixelFormat::RGB16Snorm)
.value("RGBA16_SNORM", PixelFormat::RGBA16Snorm)
python: bind PixelFormat and PixelStorage. Just the uncompressed ones for now.
7 years ago
.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)
python: expose depth/stencil PixelFormat values.
4 years ago
.value("RGBA32F", PixelFormat::RGBA32F)
.value("DEPTH16_UNORM", PixelFormat::Depth16Unorm)
.value("DEPTH24_UNORM", PixelFormat::Depth24Unorm)
.value("DEPTH32F", PixelFormat::Depth32F)
.value("STENCIL8UI", PixelFormat::Stencil8UI)
.value("DEPTH16_UNORM_STENCIL8UI", PixelFormat::Depth16UnormStencil8UI)
.value("DEPTH24_UNORM_STENCIL8UI", PixelFormat::Depth24UnormStencil8UI)
.value("DEPTH32F_STENCIL8UI", PixelFormat::Depth32FStencil8UI);
python: bind PixelFormat and PixelStorage. Just the uncompressed ones for now.
7 years ago
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: expose Image. Owning counterpart to ImageView.
7 years ago
py::class_<Image1D> image1D{m, "Image1D", "One-dimensional image"};
py::class_<Image2D> image2D{m, "Image2D", "Two-dimensional image"};
py::class_<Image3D> image3D{m, "Image3D", "Three-dimensional image"};
image(image1D);
image(image2D);
image(image3D);
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: expose textures and related APIs.
7 years ago
py::enum_<SamplerFilter>{m, "SamplerFilter", "Texture sampler filtering"}
.value("NEAREST", SamplerFilter::Nearest)
.value("LINEAR", SamplerFilter::Linear);
py::enum_<SamplerMipmap>{m, "SamplerMipmap", "Texture sampler mip level selection"}
.value("BASE", SamplerMipmap::Base)
.value("NEAREST", SamplerMipmap::Nearest)
.value("LINEAR", SamplerMipmap::Linear);
py::enum_<SamplerWrapping>{m, "SamplerWrapping", "Texture sampler wrapping"}
.value("REPEAT", SamplerWrapping::Repeat)
.value("MIRRORED_REPEAT", SamplerWrapping::MirroredRepeat)
.value("CLAMP_TO_EDGE", SamplerWrapping::ClampToEdge)
.value("CLAMP_TO_BORDER", SamplerWrapping::ClampToBorder)
.value("MIRROR_CLAMP_TO_EDGE", SamplerWrapping::MirrorClampToEdge);
python: add generic MeshPrimitive and MeshIndexType enums.
7 years ago
}
}}
python: fix Clang's -Wmissing-prototypes warnings. This is a temporary change until a PR fixing this on pybind11 side is merged (and a new version released).
7 years ago
/* TODO: remove declaration when https://github.com/pybind/pybind11/pull/1863
is released */
python: fix PyInit_* declarations to work on Windows as well. Co-authored-by: Cameron Egbert <cegbert@oculus.com>
7 years ago
extern "C" PYBIND11_EXPORT PyObject* PyInit__magnum();
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: minor cleanup.
7 years ago
m.doc() = "Root Magnum module";
python: expose (Mutable)ImageView.
7 years ago
/* We need ArrayView for images */
python: make this compatible with pybind11 2.6. Pybind changed py::module to py::module_ in order to support C++ modules, adapting to that change. It still builds on older versions, but we're using only the new APIs.
6 years ago
py::module_::import("corrade.containers");
python: expose (Mutable)ImageView.
7 years ago
python: make this compatible with pybind11 2.6. Pybind changed py::module to py::module_ in order to support C++ modules, adapting to that change. It still builds on older versions, but we're using only the new APIs.
6 years ago
py::module_ math = m.def_submodule("math");
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
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
python: make this compatible with pybind11 2.6. Pybind changed py::module to py::module_ in order to support C++ modules, adapting to that change. It still builds on older versions, but we're using only the new APIs.
6 years ago
py::module_ gl = m.def_submodule("gl");
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
magnum::gl(gl);
#endif
#ifdef Magnum_SceneGraph_FOUND
python: make this compatible with pybind11 2.6. Pybind changed py::module to py::module_ in order to support C++ modules, adapting to that change. It still builds on older versions, but we're using only the new APIs.
6 years ago
py::module_ scenegraph = m.def_submodule("scenegraph");
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
magnum::scenegraph(scenegraph);
#endif
python: expose the essentials from the Text library. Everything needed to make a Python variant of the Text example. The Oxygen.ttf font file is the same as in MagnumPlugins/FreeTypeFont/Test in the magnum-plugins repository.
4 years ago
#ifdef Magnum_Text_FOUND
py::module_ text = m.def_submodule("text");
magnum::text(text);
#endif
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_Trade_FOUND
python: make this compatible with pybind11 2.6. Pybind changed py::module to py::module_ in order to support C++ modules, adapting to that change. It still builds on older versions, but we're using only the new APIs.
6 years ago
py::module_ trade = m.def_submodule("trade");
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
magnum::trade(trade);
#endif
#ifdef Magnum_MeshTools_FOUND
/* Depends on trade and gl */
python: make this compatible with pybind11 2.6. Pybind changed py::module to py::module_ in order to support C++ modules, adapting to that change. It still builds on older versions, but we're using only the new APIs.
6 years ago
py::module_ meshtools = m.def_submodule("meshtools");
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
magnum::meshtools(meshtools);
#endif
#ifdef Magnum_Primitives_FOUND
/* Depends on trade */
python: make this compatible with pybind11 2.6. Pybind changed py::module to py::module_ in order to support C++ modules, adapting to that change. It still builds on older versions, but we're using only the new APIs.
6 years ago
py::module_ primitives = m.def_submodule("primitives");
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
magnum::primitives(primitives);
#endif
#ifdef Magnum_Shaders_FOUND
/* Depends on gl */
python: make this compatible with pybind11 2.6. Pybind changed py::module to py::module_ in order to support C++ modules, adapting to that change. It still builds on older versions, but we're using only the new APIs.
6 years ago
py::module_ shaders = m.def_submodule("shaders");
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
magnum::shaders(shaders);
#endif
/* Keep the doc in sync with platform/__init__.py */
python: make this compatible with pybind11 2.6. Pybind changed py::module to py::module_ in order to support C++ modules, adapting to that change. It still builds on older versions, but we're using only the new APIs.
6 years ago
py::module_ platform = m.def_submodule("platform");
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
platform.doc() = "Platform-specific application and context creation";
#ifdef Magnum_GlfwApplication_FOUND
python: make this compatible with pybind11 2.6. Pybind changed py::module to py::module_ in order to support C++ modules, adapting to that change. It still builds on older versions, but we're using only the new APIs.
6 years ago
py::module_ glfw = platform.def_submodule("glfw");
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
magnum::platform::glfw(glfw);
#endif
#ifdef Magnum_Sdl2Application_FOUND
python: make this compatible with pybind11 2.6. Pybind changed py::module to py::module_ in order to support C++ modules, adapting to that change. It still builds on older versions, but we're using only the new APIs.
6 years ago
py::module_ sdl2 = platform.def_submodule("sdl2");
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
magnum::platform::sdl2(sdl2);
#endif
#ifdef Magnum_WindowlessEglApplication_FOUND
python: make this compatible with pybind11 2.6. Pybind changed py::module to py::module_ in order to support C++ modules, adapting to that change. It still builds on older versions, but we're using only the new APIs.
6 years ago
py::module_ egl = platform.def_submodule("egl");
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
magnum::platform::egl(egl);
#endif
#ifdef Magnum_WindowlessGlxApplication_FOUND
python: make this compatible with pybind11 2.6. Pybind changed py::module to py::module_ in order to support C++ modules, adapting to that change. It still builds on older versions, but we're using only the new APIs.
6 years ago
py::module_ glx = platform.def_submodule("glx");
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
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
}