Zengtudor/pybind11
1
0
Fork 0
mirror of https://github.com/pybind/pybind11.git synced 2025-02-17 06:00:51 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add doc of py::numpy_scalar

Browse Source
This commit is contained in:
sun1638650145 2021-12-07 11:22:45 +08:00
parent 58771146a9
commit 9d846f6e27
2 changed files with 44 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

40
docs/advanced/pycpp/numpy.rst
View File

@ -232,6 +232,46 @@ prevent many types of unsupported structures, it is still the user's
responsibility to use only "plain" structures that can be safely manipulated as
raw memory without violating invariants.
Scalar types
============
In some cases we may want to accept or return NumPy scalar values such as
``np.float32`` or ``np.float64``. We hope to be able to handle single-precision
and double-precision on the C-side. However, both are bound to Python's
double-precision builtin float by default, so they cannot be processed separately.
We used the ``py::buffer`` trick to implement the previous approach, which
will cause the readability of the code to drop significantly.
Luckily, there's a helper type for this occasion - ``py::numpy_scalar``:
.. code-block:: cpp
m.def("add", [](py::numpy_scalar<float> a, py::numpy_scalar<float> b) {
return py::make_scalar(a + b);
});
m.def("add", [](py::numpy_scalar<double> a, py::numpy_scalar<double> b) {
return py::make_scalar(a + b);
});
This type is trivially convertible to and from the type it wraps; currently
supported scalar types are NumPy arithmetic types: ``bool_``, ``int8``,
``int16``, ``int32``, ``int64``, ``uint8``, ``uint16``, ``uint32``,
``uint64``, ``float32``, ``float64``, ``complex64``, ``complex128``, all of
them mapping to respective C++ counterparts.
.. note::
This is a strict type, it will only allow input arguments of the specified
NumPy type and nothing else (e.g., ``py::numpy_scalar<int64_t>`` will not
accept built-in ``int`` or any other type for that matter).
.. note::
Native C types are mapped to NumPy types in a platform specific way: for
instance, ``char`` may be mapped to either ``np.int8`` or ``np.uint8``
depending on the platform. If you want to ensure specific NumPy types,
it is recommended to use fixed-width aliases from ``<cstdint>``.
Vectorizing functions
=====================

8
include/pybind11/numpy.h
View File

@ -9,8 +9,8 @@
#pragma once
#include "pybind11.h"
#include "complex.h"
#include "pybind11/pybind11.h"
#include "pybind11/complex.h"
#include <numeric>
#include <algorithm>
#include <array>
@ -159,9 +159,9 @@ struct npy_api {
NPY_DOUBLE_, NPY_FLOAT_, NPY_LONGDOUBLE_),
NPY_FLOAT64_ = platform_lookup<8, double, float, long double>(
NPY_DOUBLE_, NPY_FLOAT_, NPY_LONGDOUBLE_),
NPY_COMPLEX64_ = platform_lookup<8, double, float, long double>(
NPY_COMPLEX64_ = platform_lookup<8, std::complex<double>, std::complex<float>, std::complex<long double>>(
NPY_DOUBLE_, NPY_FLOAT_, NPY_LONGDOUBLE_),
NPY_COMPLEX128_ = platform_lookup<8, double, float, long double>(
NPY_COMPLEX128_ = platform_lookup<8, std::complex<double>, std::complex<float>, std::complex<long double>>(
NPY_DOUBLE_, NPY_FLOAT_, NPY_LONGDOUBLE_),
NPY_CHAR_ = std::is_signed<char>::value ? NPY_BYTE_ : NPY_UBYTE_,
};