github.com/kaydxh/golang@v0.0.131/pkg/gocv/cgo/third_path/pybind11/docs/advanced/smart_ptrs.rst (about)

     1  Smart pointers
     2  ##############
     3  
     4  std::unique_ptr
     5  ===============
     6  
     7  Given a class ``Example`` with Python bindings, it's possible to return
     8  instances wrapped in C++11 unique pointers, like so
     9  
    10  .. code-block:: cpp
    11  
    12      std::unique_ptr<Example> create_example() { return std::unique_ptr<Example>(new Example()); }
    13  
    14  .. code-block:: cpp
    15  
    16      m.def("create_example", &create_example);
    17  
    18  In other words, there is nothing special that needs to be done. While returning
    19  unique pointers in this way is allowed, it is *illegal* to use them as function
    20  arguments. For instance, the following function signature cannot be processed
    21  by pybind11.
    22  
    23  .. code-block:: cpp
    24  
    25      void do_something_with_example(std::unique_ptr<Example> ex) { ... }
    26  
    27  The above signature would imply that Python needs to give up ownership of an
    28  object that is passed to this function, which is generally not possible (for
    29  instance, the object might be referenced elsewhere).
    30  
    31  std::shared_ptr
    32  ===============
    33  
    34  The binding generator for classes, :class:`class_`, can be passed a template
    35  type that denotes a special *holder* type that is used to manage references to
    36  the object.  If no such holder type template argument is given, the default for
    37  a type named ``Type`` is ``std::unique_ptr<Type>``, which means that the object
    38  is deallocated when Python's reference count goes to zero.
    39  
    40  It is possible to switch to other types of reference counting wrappers or smart
    41  pointers, which is useful in codebases that rely on them. For instance, the
    42  following snippet causes ``std::shared_ptr`` to be used instead.
    43  
    44  .. code-block:: cpp
    45  
    46      py::class_<Example, std::shared_ptr<Example> /* <- holder type */> obj(m, "Example");
    47  
    48  Note that any particular class can only be associated with a single holder type.
    49  
    50  One potential stumbling block when using holder types is that they need to be
    51  applied consistently. Can you guess what's broken about the following binding
    52  code?
    53  
    54  .. code-block:: cpp
    55  
    56      class Child { };
    57  
    58      class Parent {
    59      public:
    60         Parent() : child(std::make_shared<Child>()) { }
    61         Child *get_child() { return child.get(); }  /* Hint: ** DON'T DO THIS ** */
    62      private:
    63          std::shared_ptr<Child> child;
    64      };
    65  
    66      PYBIND11_MODULE(example, m) {
    67          py::class_<Child, std::shared_ptr<Child>>(m, "Child");
    68  
    69          py::class_<Parent, std::shared_ptr<Parent>>(m, "Parent")
    70             .def(py::init<>())
    71             .def("get_child", &Parent::get_child);
    72      }
    73  
    74  The following Python code will cause undefined behavior (and likely a
    75  segmentation fault).
    76  
    77  .. code-block:: python
    78  
    79     from example import Parent
    80  
    81     print(Parent().get_child())
    82  
    83  The problem is that ``Parent::get_child()`` returns a pointer to an instance of
    84  ``Child``, but the fact that this instance is already managed by
    85  ``std::shared_ptr<...>`` is lost when passing raw pointers. In this case,
    86  pybind11 will create a second independent ``std::shared_ptr<...>`` that also
    87  claims ownership of the pointer. In the end, the object will be freed **twice**
    88  since these shared pointers have no way of knowing about each other.
    89  
    90  There are two ways to resolve this issue:
    91  
    92  1. For types that are managed by a smart pointer class, never use raw pointers
    93     in function arguments or return values. In other words: always consistently
    94     wrap pointers into their designated holder types (such as
    95     ``std::shared_ptr<...>``). In this case, the signature of ``get_child()``
    96     should be modified as follows:
    97  
    98  .. code-block:: cpp
    99  
   100      std::shared_ptr<Child> get_child() { return child; }
   101  
   102  2. Adjust the definition of ``Child`` by specifying
   103     ``std::enable_shared_from_this<T>`` (see cppreference_ for details) as a
   104     base class. This adds a small bit of information to ``Child`` that allows
   105     pybind11 to realize that there is already an existing
   106     ``std::shared_ptr<...>`` and communicate with it. In this case, the
   107     declaration of ``Child`` should look as follows:
   108  
   109  .. _cppreference: http://en.cppreference.com/w/cpp/memory/enable_shared_from_this
   110  
   111  .. code-block:: cpp
   112  
   113      class Child : public std::enable_shared_from_this<Child> { };
   114  
   115  .. _smart_pointers:
   116  
   117  Custom smart pointers
   118  =====================
   119  
   120  pybind11 supports ``std::unique_ptr`` and ``std::shared_ptr`` right out of the
   121  box. For any other custom smart pointer, transparent conversions can be enabled
   122  using a macro invocation similar to the following. It must be declared at the
   123  top namespace level before any binding code:
   124  
   125  .. code-block:: cpp
   126  
   127      PYBIND11_DECLARE_HOLDER_TYPE(T, SmartPtr<T>);
   128  
   129  The first argument of :func:`PYBIND11_DECLARE_HOLDER_TYPE` should be a
   130  placeholder name that is used as a template parameter of the second argument.
   131  Thus, feel free to use any identifier, but use it consistently on both sides;
   132  also, don't use the name of a type that already exists in your codebase.
   133  
   134  The macro also accepts a third optional boolean parameter that is set to false
   135  by default. Specify
   136  
   137  .. code-block:: cpp
   138  
   139      PYBIND11_DECLARE_HOLDER_TYPE(T, SmartPtr<T>, true);
   140  
   141  if ``SmartPtr<T>`` can always be initialized from a ``T*`` pointer without the
   142  risk of inconsistencies (such as multiple independent ``SmartPtr`` instances
   143  believing that they are the sole owner of the ``T*`` pointer). A common
   144  situation where ``true`` should be passed is when the ``T`` instances use
   145  *intrusive* reference counting.
   146  
   147  Please take a look at the :ref:`macro_notes` before using this feature.
   148  
   149  By default, pybind11 assumes that your custom smart pointer has a standard
   150  interface, i.e. provides a ``.get()`` member function to access the underlying
   151  raw pointer. If this is not the case, pybind11's ``holder_helper`` must be
   152  specialized:
   153  
   154  .. code-block:: cpp
   155  
   156      // Always needed for custom holder types
   157      PYBIND11_DECLARE_HOLDER_TYPE(T, SmartPtr<T>);
   158  
   159      // Only needed if the type's `.get()` goes by another name
   160      namespace PYBIND11_NAMESPACE { namespace detail {
   161          template <typename T>
   162          struct holder_helper<SmartPtr<T>> { // <-- specialization
   163              static const T *get(const SmartPtr<T> &p) { return p.getPointer(); }
   164          };
   165      }}
   166  
   167  The above specialization informs pybind11 that the custom ``SmartPtr`` class
   168  provides ``.get()`` functionality via ``.getPointer()``.
   169  
   170  .. seealso::
   171  
   172      The file :file:`tests/test_smart_ptr.cpp` contains a complete example
   173      that demonstrates how to work with custom reference-counting holder types
   174      in more detail.