2. Defining New Types

As mentioned in the last chapter, Python allows the writer of an extension module to define new types that can be manipulated from Python code, much like strings and lists in core Python.

This is not hard; the code for all extension types follows a pattern, but there are some details that you need to understand before you can get started.

2.1. The Basics

The Python runtime sees all Python objects as variables of type PyObject*, which serves as a "base type" for all Python objects. PyObject itself only contains the refcount and a pointer to the object's "type object". This is where the action is; the type object determines which (C) functions get called when, for instance, an attribute gets looked up on an object or it is multiplied by another object. These C functions are called "type methods".

So, if you want to define a new object type, you need to create a new type object.

This sort of thing can only be explained by example, so here's a minimal, but complete, module that defines a new type:

#include <Python.h>

typedef struct {
    PyObject_HEAD
    /* Type-specific fields go here. */
} noddy_NoddyObject;

static PyTypeObject noddy_NoddyType = {
    PyVarObject_HEAD_INIT(NULL, 0)
    "noddy.Noddy",             /* tp_name */
    sizeof(noddy_NoddyObject), /* tp_basicsize */
    0,                         /* tp_itemsize */
    0,                         /* tp_dealloc */
    0,                         /* tp_print */
    0,                         /* tp_getattr */
    0,                         /* tp_setattr */
    0,                         /* tp_reserved */
    0,                         /* tp_repr */
    0,                         /* tp_as_number */
    0,                         /* tp_as_sequence */
    0,                         /* tp_as_mapping */
    0,                         /* tp_hash  */
    0,                         /* tp_call */
    0,                         /* tp_str */
    0,                         /* tp_getattro */
    0,                         /* tp_setattro */
    0,                         /* tp_as_buffer */
    Py_TPFLAGS_DEFAULT,        /* tp_flags */
    "Noddy objects",           /* tp_doc */
};

static PyModuleDef noddymodule = {
    PyModuleDef_HEAD_INIT,
    "noddy",
    "Example module that creates an extension type.",
    -1,
    NULL, NULL, NULL, NULL, NULL
};

PyMODINIT_FUNC
PyInit_noddy(void)
{
    PyObject* m;

    noddy_NoddyType.tp_new = PyType_GenericNew;
    if (PyType_Ready(&noddy_NoddyType) < 0)
        return NULL;

    m = PyModule_Create(&noddymodule);
    if (m == NULL)
        return NULL;

    Py_INCREF(&noddy_NoddyType);
    PyModule_AddObject(m, "Noddy", (PyObject *)&noddy_NoddyType);
    return m;
}

Now that's quite a bit to take in at once, but hopefully bits will seem familiar from the last chapter.

The first bit that will be new is:

typedef struct {
    PyObject_HEAD
} noddy_NoddyObject;

This is what a Noddy object will contain---in this case, nothing more than what every Python object contains---a field called ob_base of type PyObject. PyObject in turn, contains an ob_refcnt field and a pointer to a type object. These can be accessed using the macros Py_REFCNT and Py_TYPE respectively. These are the fields the PyObject_HEAD macro brings in. The reason for the macro is to standardize the layout and to enable special debugging fields in debug builds.

Note that there is no semicolon after the PyObject_HEAD macro; one is included in the macro definition. Be wary of adding one by accident; it's easy to do from habit, and your compiler might not complain, but someone else's probably will! (On Windows, MSVC is known to call this an error and refuse to compile the code.)

For contrast, let's take a look at the corresponding definition for standard Python floats:

typedef struct {
    PyObject_HEAD
    double ob_fval;
} PyFloatObject;

Moving on, we come to the crunch --- the type object.

static PyTypeObject noddy_NoddyType = {
    PyVarObject_HEAD_INIT(NULL, 0)
    "noddy.Noddy",             /* tp_name */
    sizeof(noddy_NoddyObject), /* tp_basicsize */
    0,                         /* tp_itemsize */
    0,                         /* tp_dealloc */
    0,                         /* tp_print */
    0,                         /* tp_getattr */
    0,                         /* tp_setattr */
    0,                         /* tp_as_async */
    0,                         /* tp_repr */
    0,                         /* tp_as_number */
    0,                         /* tp_as_sequence */
    0,                         /* tp_as_mapping */
    0,                         /* tp_hash  */
    0,                         /* tp_call */
    0,                         /* tp_str */
    0,                         /* tp_getattro */
    0,                         /* tp_setattro */
    0,                         /* tp_as_buffer */
    Py_TPFLAGS_DEFAULT,        /* tp_flags */
    "Noddy objects",           /* tp_doc */
};

Now if you go and look up the definition of PyTypeObject in object.h you'll see that it has many more fields that the definition above. The remaining fields will be filled with zeros by the C compiler, and it's common practice to not specify them explicitly unless you need them.

This is so important that we're going to pick the top of it apart still further:

PyVarObject_HEAD_INIT(NULL, 0)

This line is a bit of a wart; what we'd like to write is:

PyVarObject_HEAD_INIT(&PyType_Type, 0)

as the type of a type object is "type", but this isn't strictly conforming C and some compilers complain. Fortunately, this member will be filled in for us by PyType_Ready().

"noddy.Noddy",              /* tp_name */

The name of our type. This will appear in the default textual representation of our objects and in some error messages, for example:

>>> "" + noddy.new_noddy()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: cannot add type "noddy.Noddy" to string

Note that the name is a dotted name that includes both the module name and the name of the type within the module. The module in this case is noddy and the type is Noddy, so we set the type name to noddy.Noddy. One side effect of using an undotted name is that the pydoc documentation tool will not list the new type in the module documentation.

sizeof(noddy_NoddyObject),  /* tp_basicsize */

This is so that Python knows how much memory to allocate when you call PyObject_New().

注釈

If you want your type to be subclassable from Python, and your type has the same tp_basicsize as its base type, you may have problems with multiple inheritance. A Python subclass of your type will have to list your type first in its __bases__, or else it will not be able to call your type's __new__() method without getting an error. You can avoid this problem by ensuring that your type has a larger value for tp_basicsize than its base type does. Most of the time, this will be true anyway, because either your base type will be object, or else you will be adding data members to your base type, and therefore increasing its size.

0,                          /* tp_itemsize */

This has to do with variable length objects like lists and strings. Ignore this for now.

Skipping a number of type methods that we don't provide, we set the class flags to Py_TPFLAGS_DEFAULT.

Py_TPFLAGS_DEFAULT,        /* tp_flags */

All types should include this constant in their flags. It enables all of the members defined until at least Python 3.3. If you need further members, you will need to OR the corresponding flags.

We provide a doc string for the type in tp_doc.

"Noddy objects",           /* tp_doc */

Now we get into the type methods, the things that make your objects different from the others. We aren't going to implement any of these in this version of the module. We'll expand this example later to have more interesting behavior.

For now, all we want to be able to do is to create new Noddy objects. To enable object creation, we have to provide a tp_new implementation. In this case, we can just use the default implementation provided by the API function PyType_GenericNew(). We'd like to just assign this to the tp_new slot, but we can't, for portability sake, On some platforms or compilers, we can't statically initialize a structure member with a function defined in another C module, so, instead, we'll assign the tp_new slot in the module initialization function just before calling PyType_Ready():

noddy_NoddyType.tp_new = PyType_GenericNew;
if (PyType_Ready(&noddy_NoddyType) < 0)
    return;

All the other type methods are NULL, so we'll go over them later --- that's for a later section!

Everything else in the file should be familiar, except for some code in PyInit_noddy():

if (PyType_Ready(&noddy_NoddyType) < 0)
    return;

This initializes the Noddy type, filing in a number of members, including ob_type that we initially set to NULL.

PyModule_AddObject(m, "Noddy", (PyObject *)&noddy_NoddyType);

This adds the type to the module dictionary. This allows us to create Noddy instances by calling the Noddy class:

>>> import noddy
>>> mynoddy = noddy.Noddy()

That's it! All that remains is to build it; put the above code in a file called noddy.c and

from distutils.core import setup, Extension
setup(name="noddy", version="1.0",
      ext_modules=[Extension("noddy", ["noddy.c"])])

in a file called setup.py; then typing

$ python setup.py build

at a shell should produce a file noddy.so in a subdirectory; move to that directory and fire up Python --- you should be able to import noddy and play around with Noddy objects.

That wasn't so hard, was it?

Of course, the current Noddy type is pretty uninteresting. It has no data and doesn't do anything. It can't even be subclassed.

2.1.1. Adding data and methods to the Basic example

Let's extend the basic example to add some data and methods. Let's also make the type usable as a base class. We'll create a new module, noddy2 that adds these capabilities:

#include <Python.h>
#include "structmember.h"

typedef struct {
    PyObject_HEAD
    PyObject *first; /* first name */
    PyObject *last;  /* last name */
    int number;
} Noddy;

static void
Noddy_dealloc(Noddy* self)
{
    Py_XDECREF(self->first);
    Py_XDECREF(self->last);
    Py_TYPE(self)->tp_free((PyObject*)self);
}

static PyObject *
Noddy_new(PyTypeObject *type, PyObject *args, PyObject *kwds)
{
    Noddy *self;

    self = (Noddy *)type->tp_alloc(type, 0);
    if (self != NULL) {
        self->first = PyUnicode_FromString("");
        if (self->first == NULL) {
            Py_DECREF(self);
            return NULL;
        }

        self->last = PyUnicode_FromString("");
        if (self->last == NULL) {
            Py_DECREF(self);
            return NULL;
        }

        self->number = 0;
    }

    return (PyObject *)self;
}

static int
Noddy_init(Noddy *self, PyObject *args, PyObject *kwds)
{
    PyObject *first=NULL, *last=NULL, *tmp;

    static char *kwlist[] = {"first", "last", "number", NULL};

    if (! PyArg_ParseTupleAndKeywords(args, kwds, "|OOi", kwlist,
                                      &first, &last,
                                      &self->number))
        return -1;

    if (first) {
        tmp = self->first;
        Py_INCREF(first);
        self->first = first;
        Py_XDECREF(tmp);
    }

    if (last) {
        tmp = self->last;
        Py_INCREF(last);
        self->last = last;
        Py_XDECREF(tmp);
    }

    return 0;
}


static PyMemberDef Noddy_members[] = {
    {"first", T_OBJECT_EX, offsetof(Noddy, first), 0,
     "first name"},
    {"last", T_OBJECT_EX, offsetof(Noddy, last), 0,
     "last name"},
    {"number", T_INT, offsetof(Noddy, number), 0,
     "noddy number"},
    {NULL}  /* Sentinel */
};

static PyObject *
Noddy_name(Noddy* self)
{
    if (self->first == NULL) {
        PyErr_SetString(PyExc_AttributeError, "first");
        return NULL;
    }

    if (self->last == NULL) {
        PyErr_SetString(PyExc_AttributeError, "last");
        return NULL;
    }

    return PyUnicode_FromFormat("%S %S", self->first, self->last);
}

static PyMethodDef Noddy_methods[] = {
    {"name", (PyCFunction)Noddy_name, METH_NOARGS,
     "Return the name, combining the first and last name"
    },
    {NULL}  /* Sentinel */
};

static PyTypeObject NoddyType = {
    PyVarObject_HEAD_INIT(NULL, 0)
    "noddy.Noddy",             /* tp_name */
    sizeof(Noddy),             /* tp_basicsize */
    0,                         /* tp_itemsize */
    (destructor)Noddy_dealloc, /* tp_dealloc */
    0,                         /* tp_print */
    0,                         /* tp_getattr */
    0,                         /* tp_setattr */
    0,                         /* tp_reserved */
    0,                         /* tp_repr */
    0,                         /* tp_as_number */
    0,                         /* tp_as_sequence */
    0,                         /* tp_as_mapping */
    0,                         /* tp_hash  */
    0,                         /* tp_call */
    0,                         /* tp_str */
    0,                         /* tp_getattro */
    0,                         /* tp_setattro */
    0,                         /* tp_as_buffer */
    Py_TPFLAGS_DEFAULT |
        Py_TPFLAGS_BASETYPE,   /* tp_flags */
    "Noddy objects",           /* tp_doc */
    0,                         /* tp_traverse */
    0,                         /* tp_clear */
    0,                         /* tp_richcompare */
    0,                         /* tp_weaklistoffset */
    0,                         /* tp_iter */
    0,                         /* tp_iternext */
    Noddy_methods,             /* tp_methods */
    Noddy_members,             /* tp_members */
    0,                         /* tp_getset */
    0,                         /* tp_base */
    0,                         /* tp_dict */
    0,                         /* tp_descr_get */
    0,                         /* tp_descr_set */
    0,                         /* tp_dictoffset */
    (initproc)Noddy_init,      /* tp_init */
    0,                         /* tp_alloc */
    Noddy_new,                 /* tp_new */
};

static PyModuleDef noddy2module = {
    PyModuleDef_HEAD_INIT,
    "noddy2",
    "Example module that creates an extension type.",
    -1,
    NULL, NULL, NULL, NULL, NULL
};

PyMODINIT_FUNC
PyInit_noddy2(void)
{
    PyObject* m;

    if (PyType_Ready(&NoddyType) < 0)
        return NULL;

    m = PyModule_Create(&noddy2module);
    if (m == NULL)
        return NULL;

    Py_INCREF(&NoddyType);
    PyModule_AddObject(m, "Noddy", (PyObject *)&NoddyType);
    return m;
}

This version of the module has a number of changes.

We've added an extra include:

#include <structmember.h>

This include provides declarations that we use to handle attributes, as described a bit later.

The name of the Noddy object structure has been shortened to Noddy. The type object name has been shortened to NoddyType.

The Noddy type now has three data attributes, first, last, and number. The first and last variables are Python strings containing first and last names. The number attribute is an integer.

The object structure is updated accordingly:

typedef struct {
    PyObject_HEAD
    PyObject *first;
    PyObject *last;
    int number;
} Noddy;

Because we now have data to manage, we have to be more careful about object allocation and deallocation. At a minimum, we need a deallocation method:

static void
Noddy_dealloc(Noddy* self)
{
    Py_XDECREF(self->first);
    Py_XDECREF(self->last);
    Py_TYPE(self)->tp_free((PyObject*)self);
}

which is assigned to the tp_dealloc member:

(destructor)Noddy_dealloc, /*tp_dealloc*/

This method decrements the reference counts of the two Python attributes. We use Py_XDECREF() here because the first and last members could be NULL. It then calls the tp_free member of the object's type to free the object's memory. Note that the object's type might not be NoddyType, because the object may be an instance of a subclass.

We want to make sure that the first and last names are initialized to empty strings, so we provide a new method:

static PyObject *
Noddy_new(PyTypeObject *type, PyObject *args, PyObject *kwds)
{
    Noddy *self;

    self = (Noddy *)type->tp_alloc(type, 0);
    if (self != NULL) {
        self->first = PyUnicode_FromString("");
        if (self->first == NULL) {
            Py_DECREF(self);
            return NULL;
        }

        self->last = PyUnicode_FromString("");
        if (self->last == NULL) {
            Py_DECREF(self);
            return NULL;
        }

        self->number = 0;
    }

    return (PyObject *)self;
}

and install it in the tp_new member:

Noddy_new,                 /* tp_new */

The new member is responsible for creating (as opposed to initializing) objects of the type. It is exposed in Python as the __new__() method. See the paper titled "Unifying types and classes in Python" for a detailed discussion of the __new__() method. One reason to implement a new method is to assure the initial values of instance variables. In this case, we use the new method to make sure that the initial values of the members first and last are not NULL. If we didn't care whether the initial values were NULL, we could have used PyType_GenericNew() as our new method, as we did before. PyType_GenericNew() initializes all of the instance variable members to NULL.

The new method is a static method that is passed the type being instantiated and any arguments passed when the type was called, and that returns the new object created. New methods always accept positional and keyword arguments, but they often ignore the arguments, leaving the argument handling to initializer methods. Note that if the type supports subclassing, the type passed may not be the type being defined. The new method calls the tp_alloc slot to allocate memory. We don't fill the tp_alloc slot ourselves. Rather PyType_Ready() fills it for us by inheriting it from our base class, which is object by default. Most types use the default allocation.

注釈

If you are creating a co-operative tp_new (one that calls a base type's tp_new or __new__()), you must not try to determine what method to call using method resolution order at runtime. Always statically determine what type you are going to call, and call its tp_new directly, or via type->tp_base->tp_new. If you do not do this, Python subclasses of your type that also inherit from other Python-defined classes may not work correctly. (Specifically, you may not be able to create instances of such subclasses without getting a TypeError.)

We provide an initialization function:

static int
Noddy_init(Noddy *self, PyObject *args, PyObject *kwds)
{
    PyObject *first=NULL, *last=NULL, *tmp;

    static char *kwlist[] = {"first", "last", "number", NULL};

    if (! PyArg_ParseTupleAndKeywords(args, kwds, "|OOi", kwlist,
                                      &first, &last,
                                      &self->number))
        return -1;

    if (first) {
        tmp = self->first;
        Py_INCREF(first);
        self->first = first;
        Py_XDECREF(tmp);
    }

    if (last) {
        tmp = self->last;
        Py_INCREF(last);
        self->last = last;
        Py_XDECREF(tmp);
    }

    return 0;
}

by filling the tp_init slot.

(initproc)Noddy_init,         /* tp_init */

The tp_init slot is exposed in Python as the __init__() method. It is used to initialize an object after it's created. Unlike the new method, we can't guarantee that the initializer is called. The initializer isn't called when unpickling objects and it can be overridden. Our initializer accepts arguments to provide initial values for our instance. Initializers always accept positional and keyword arguments. Initializers should return either 0 on success or -1 on error.

Initializers can be called multiple times. Anyone can call the __init__() method on our objects. For this reason, we have to be extra careful when assigning the new values. We might be tempted, for example to assign the first member like this:

if (first) {
    Py_XDECREF(self->first);
    Py_INCREF(first);
    self->first = first;
}

But this would be risky. Our type doesn't restrict the type of the first member, so it could be any kind of object. It could have a destructor that causes code to be executed that tries to access the first member. To be paranoid and protect ourselves against this possibility, we almost always reassign members before decrementing their reference counts. When don't we have to do this?

• when we absolutely know that the reference count is greater than 1
• when we know that deallocation of the object [1] will not cause any calls back into our type's code
• when decrementing a reference count in a tp_dealloc handler when garbage-collections is not supported [2]

We want to expose our instance variables as attributes. There are a number of ways to do that. The simplest way is to define member definitions:

static PyMemberDef Noddy_members[] = {
    {"first", T_OBJECT_EX, offsetof(Noddy, first), 0,
     "first name"},
    {"last", T_OBJECT_EX, offsetof(Noddy, last), 0,
     "last name"},
    {"number", T_INT, offsetof(Noddy, number), 0,
     "noddy number"},
    {NULL}  /* Sentinel */
};

and put the definitions in the tp_members slot:

Noddy_members,             /* tp_members */

Each member definition has a member name, type, offset, access flags and documentation string. See the 総称的な属性を管理する section below for details.

A disadvantage of this approach is that it doesn't provide a way to restrict the types of objects that can be assigned to the Python attributes. We expect the first and last names to be strings, but any Python objects can be assigned. Further, the attributes can be deleted, setting the C pointers to NULL. Even though we can make sure the members are initialized to non-NULL values, the members can be set to NULL if the attributes are deleted.

We define a single method, name(), that outputs the objects name as the concatenation of the first and last names.

static PyObject *
Noddy_name(Noddy* self)
{
    if (self->first == NULL) {
        PyErr_SetString(PyExc_AttributeError, "first");
        return NULL;
    }

    if (self->last == NULL) {
        PyErr_SetString(PyExc_AttributeError, "last");
        return NULL;
    }

    return PyUnicode_FromFormat("%S %S", self->first, self->last);
}

The method is implemented as a C function that takes a Noddy (or Noddy subclass) instance as the first argument. Methods always take an instance as the first argument. Methods often take positional and keyword arguments as well, but in this case we don't take any and don't need to accept a positional argument tuple or keyword argument dictionary. This method is equivalent to the Python method:

def name(self):
   return "%s %s" % (self.first, self.last)

Note that we have to check for the possibility that our first and last members are NULL. This is because they can be deleted, in which case they are set to NULL. It would be better to prevent deletion of these attributes and to restrict the attribute values to be strings. We'll see how to do that in the next section.

Now that we've defined the method, we need to create an array of method definitions:

static PyMethodDef Noddy_methods[] = {
    {"name", (PyCFunction)Noddy_name, METH_NOARGS,
     "Return the name, combining the first and last name"
    },
    {NULL}  /* Sentinel */
};

and assign them to the tp_methods slot:

Noddy_methods,             /* tp_methods */

Note that we used the METH_NOARGS flag to indicate that the method is passed no arguments.

Finally, we'll make our type usable as a base class. We've written our methods carefully so far so that they don't make any assumptions about the type of the object being created or used, so all we need to do is to add the Py_TPFLAGS_BASETYPE to our class flag definition:

Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE, /*tp_flags*/

We rename PyInit_noddy() to PyInit_noddy2() and update the module name in the PyModuleDef struct.

Finally, we update our setup.py file to build the new module:

from distutils.core import setup, Extension
setup(name="noddy", version="1.0",
      ext_modules=[
         Extension("noddy", ["noddy.c"]),
         Extension("noddy2", ["noddy2.c"]),
         ])

2.1.2. Providing finer control over data attributes

In this section, we'll provide finer control over how the first and last attributes are set in the Noddy example. In the previous version of our module, the instance variables first and last could be set to non-string values or even deleted. We want to make sure that these attributes always contain strings.

#include <Python.h>
#include "structmember.h"

typedef struct {
    PyObject_HEAD
    PyObject *first;
    PyObject *last;
    int number;
} Noddy;

static void
Noddy_dealloc(Noddy* self)
{
    Py_XDECREF(self->first);
    Py_XDECREF(self->last);
    Py_TYPE(self)->tp_free((PyObject*)self);
}

static PyObject *
Noddy_new(PyTypeObject *type, PyObject *args, PyObject *kwds)
{
    Noddy *self;

    self = (Noddy *)type->tp_alloc(type, 0);
    if (self != NULL) {
        self->first = PyUnicode_FromString("");
        if (self->first == NULL) {
            Py_DECREF(self);
            return NULL;
        }

        self->last = PyUnicode_FromString("");
        if (self->last == NULL) {
            Py_DECREF(self);
            return NULL;
        }

        self->number = 0;
    }

    return (PyObject *)self;
}

static int
Noddy_init(Noddy *self, PyObject *args, PyObject *kwds)
{
    PyObject *first=NULL, *last=NULL, *tmp;

    static char *kwlist[] = {"first", "last", "number", NULL};

    if (! PyArg_ParseTupleAndKeywords(args, kwds, "|SSi", kwlist,
                                      &first, &last,
                                      &self->number))
        return -1;

    if (first) {
        tmp = self->first;
        Py_INCREF(first);
        self->first = first;
        Py_DECREF(tmp);
    }

    if (last) {
        tmp = self->last;
        Py_INCREF(last);
        self->last = last;
        Py_DECREF(tmp);
    }

    return 0;
}

static PyMemberDef Noddy_members[] = {
    {"number", T_INT, offsetof(Noddy, number), 0,
     "noddy number"},
    {NULL}  /* Sentinel */
};

static PyObject *
Noddy_getfirst(Noddy *self, void *closure)
{
    Py_INCREF(self->first);
    return self->first;
}

static int
Noddy_setfirst(Noddy *self, PyObject *value, void *closure)
{
    if (value == NULL) {
        PyErr_SetString(PyExc_TypeError, "Cannot delete the first attribute");
        return -1;
    }

    if (! PyUnicode_Check(value)) {
        PyErr_SetString(PyExc_TypeError,
                        "The first attribute value must be a string");
        return -1;
    }

    Py_DECREF(self->first);
    Py_INCREF(value);
    self->first = value;

    return 0;
}

static PyObject *
Noddy_getlast(Noddy *self, void *closure)
{
    Py_INCREF(self->last);
    return self->last;
}

static int
Noddy_setlast(Noddy *self, PyObject *value, void *closure)
{
    if (value == NULL) {
        PyErr_SetString(PyExc_TypeError, "Cannot delete the last attribute");
        return -1;
    }

    if (! PyUnicode_Check(value)) {
        PyErr_SetString(PyExc_TypeError,
                        "The last attribute value must be a string");
        return -1;
    }

    Py_DECREF(self->last);
    Py_INCREF(value);
    self->last = value;

    return 0;
}

static PyGetSetDef Noddy_getseters[] = {
    {"first",
     (getter)Noddy_getfirst, (setter)Noddy_setfirst,
     "first name",
     NULL},
    {"last",
     (getter)Noddy_getlast, (setter)Noddy_setlast,
     "last name",
     NULL},
    {NULL}  /* Sentinel */
};

static PyObject *
Noddy_name(Noddy* self)
{
    return PyUnicode_FromFormat("%S %S", self->first, self->last);
}

static PyMethodDef Noddy_methods[] = {
    {"name", (PyCFunction)Noddy_name, METH_NOARGS,
     "Return the name, combining the first and last name"
    },
    {NULL}  /* Sentinel */
};

static PyTypeObject NoddyType = {
    PyVarObject_HEAD_INIT(NULL, 0)
    "noddy.Noddy",             /* tp_name */
    sizeof(Noddy),             /* tp_basicsize */
    0,                         /* tp_itemsize */
    (destructor)Noddy_dealloc, /* tp_dealloc */
    0,                         /* tp_print */
    0,                         /* tp_getattr */
    0,                         /* tp_setattr */
    0,                         /* tp_reserved */
    0,                         /* tp_repr */
    0,                         /* tp_as_number */
    0,                         /* tp_as_sequence */
    0,                         /* tp_as_mapping */
    0,                         /* tp_hash  */
    0,                         /* tp_call */
    0,                         /* tp_str */
    0,                         /* tp_getattro */
    0,                         /* tp_setattro */
    0,                         /* tp_as_buffer */
    Py_TPFLAGS_DEFAULT |
        Py_TPFLAGS_BASETYPE,   /* tp_flags */
    "Noddy objects",           /* tp_doc */
    0,                         /* tp_traverse */
    0,                         /* tp_clear */
    0,                         /* tp_richcompare */
    0,                         /* tp_weaklistoffset */
    0,                         /* tp_iter */
    0,                         /* tp_iternext */
    Noddy_methods,             /* tp_methods */
    Noddy_members,             /* tp_members */
    Noddy_getseters,           /* tp_getset */
    0,                         /* tp_base */
    0,                         /* tp_dict */
    0,                         /* tp_descr_get */
    0,                         /* tp_descr_set */
    0,                         /* tp_dictoffset */
    (initproc)Noddy_init,      /* tp_init */
    0,                         /* tp_alloc */
    Noddy_new,                 /* tp_new */
};

static PyModuleDef noddy3module = {
    PyModuleDef_HEAD_INIT,
    "noddy3",
    "Example module that creates an extension type.",
    -1,
    NULL, NULL, NULL, NULL, NULL
};

PyMODINIT_FUNC
PyInit_noddy3(void)
{
    PyObject* m;

    if (PyType_Ready(&NoddyType) < 0)
        return NULL;

    m = PyModule_Create(&noddy3module);
    if (m == NULL)
        return NULL;

    Py_INCREF(&NoddyType);
    PyModule_AddObject(m, "Noddy", (PyObject *)&NoddyType);
    return m;
}

To provide greater control, over the first and last attributes, we'll use custom getter and setter functions. Here are the functions for getting and setting the first attribute:

Noddy_getfirst(Noddy *self, void *closure)
{
    Py_INCREF(self->first);
    return self->first;
}

static int
Noddy_setfirst(Noddy *self, PyObject *value, void *closure)
{
  if (value == NULL) {
    PyErr_SetString(PyExc_TypeError, "Cannot delete the first attribute");
    return -1;
  }

  if (! PyUnicode_Check(value)) {
    PyErr_SetString(PyExc_TypeError,
                    "The first attribute value must be a str");
    return -1;
  }

  Py_DECREF(self->first);
  Py_INCREF(value);
  self->first = value;

  return 0;
}

The getter function is passed a Noddy object and a "closure", which is void pointer. In this case, the closure is ignored. (The closure supports an advanced usage in which definition data is passed to the getter and setter. This could, for example, be used to allow a single set of getter and setter functions that decide the attribute to get or set based on data in the closure.)

The setter function is passed the Noddy object, the new value, and the closure. The new value may be NULL, in which case the attribute is being deleted. In our setter, we raise an error if the attribute is deleted or if the attribute value is not a string.

We create an array of PyGetSetDef structures:

static PyGetSetDef Noddy_getseters[] = {
    {"first",
     (getter)Noddy_getfirst, (setter)Noddy_setfirst,
     "first name",
     NULL},
    {"last",
     (getter)Noddy_getlast, (setter)Noddy_setlast,
     "last name",
     NULL},
    {NULL}  /* Sentinel */
};

and register it in the tp_getset slot:

Noddy_getseters,           /* tp_getset */

to register our attribute getters and setters.

The last item in a PyGetSetDef structure is the closure mentioned above. In this case, we aren't using the closure, so we just pass NULL.

We also remove the member definitions for these attributes:

static PyMemberDef Noddy_members[] = {
    {"number", T_INT, offsetof(Noddy, number), 0,
     "noddy number"},
    {NULL}  /* Sentinel */
};

We also need to update the tp_init handler to only allow strings [3] to be passed:

static int
Noddy_init(Noddy *self, PyObject *args, PyObject *kwds)
{
    PyObject *first=NULL, *last=NULL, *tmp;

    static char *kwlist[] = {"first", "last", "number", NULL};

    if (! PyArg_ParseTupleAndKeywords(args, kwds, "|SSi", kwlist,
                                      &first, &last,
                                      &self->number))
        return -1;

    if (first) {
        tmp = self->first;
        Py_INCREF(first);
        self->first = first;
        Py_DECREF(tmp);
    }

    if (last) {
        tmp = self->last;
        Py_INCREF(last);
        self->last = last;
        Py_DECREF(tmp);
    }

    return 0;
}

With these changes, we can assure that the first and last members are never NULL so we can remove checks for NULL values in almost all cases. This means that most of the Py_XDECREF() calls can be converted to Py_DECREF() calls. The only place we can't change these calls is in the deallocator, where there is the possibility that the initialization of these members failed in the constructor.

We also rename the module initialization function and module name in the initialization function, as we did before, and we add an extra definition to the setup.py file.

2.1.3. Supporting cyclic garbage collection

Python has a cyclic-garbage collector that can identify unneeded objects even when their reference counts are not zero. This can happen when objects are involved in cycles. For example, consider:

>>> l = []
>>> l.append(l)
>>> del l

In this example, we create a list that contains itself. When we delete it, it still has a reference from itself. Its reference count doesn't drop to zero. Fortunately, Python's cyclic-garbage collector will eventually figure out that the list is garbage and free it.

In the second version of the Noddy example, we allowed any kind of object to be stored in the first or last attributes. [4] This means that Noddy objects can participate in cycles:

>>> import noddy2
>>> n = noddy2.Noddy()
>>> l = [n]
>>> n.first = l

This is pretty silly, but it gives us an excuse to add support for the cyclic-garbage collector to the Noddy example. To support cyclic garbage collection, types need to fill two slots and set a class flag that enables these slots:

#include <Python.h>
#include "structmember.h"

typedef struct {
    PyObject_HEAD
    PyObject *first;
    PyObject *last;
    int number;
} Noddy;

static int
Noddy_traverse(Noddy *self, visitproc visit, void *arg)
{
    int vret;

    if (self->first) {
        vret = visit(self->first, arg);
        if (vret != 0)
            return vret;
    }
    if (self->last) {
        vret = visit(self->last, arg);
        if (vret != 0)
            return vret;
    }

    return 0;
}

static int
Noddy_clear(Noddy *self)
{
    PyObject *tmp;

    tmp = self->first;
    self->first = NULL;
    Py_XDECREF(tmp);

    tmp = self->last;
    self->last = NULL;
    Py_XDECREF(tmp);

    return 0;
}

static void
Noddy_dealloc(Noddy* self)
{
    PyObject_GC_UnTrack(self);
    Noddy_clear(self);
    Py_TYPE(self)->tp_free((PyObject*)self);
}

static PyObject *
Noddy_new(PyTypeObject *type, PyObject *args, PyObject *kwds)
{
    Noddy *self;

    self = (Noddy *)type->tp_alloc(type, 0);
    if (self != NULL) {
        self->first = PyUnicode_FromString("");
        if (self->first == NULL) {
            Py_DECREF(self);
            return NULL;
        }

        self->last = PyUnicode_FromString("");
        if (self->last == NULL) {
            Py_DECREF(self);
            return NULL;
        }

        self->number = 0;
    }

    return (PyObject *)self;
}

static int
Noddy_init(Noddy *self, PyObject *args, PyObject *kwds)
{
    PyObject *first=NULL, *last=NULL, *tmp;

    static char *kwlist[] = {"first", "last", "number", NULL};

    if (! PyArg_ParseTupleAndKeywords(args, kwds, "|OOi", kwlist,
                                      &first, &last,
                                      &self->number))
        return -1;

    if (first) {
        tmp = self->first;
        Py_INCREF(first);
        self->first = first;
        Py_XDECREF(tmp);
    }

    if (last) {
        tmp = self->last;
        Py_INCREF(last);
        self->last = last;
        Py_XDECREF(tmp);
    }

    return 0;
}


static PyMemberDef Noddy_members[] = {
    {"first", T_OBJECT_EX, offsetof(Noddy, first), 0,
     "first name"},
    {"last", T_OBJECT_EX, offsetof(Noddy, last), 0,
     "last name"},
    {"number", T_INT, offsetof(Noddy, number), 0,
     "noddy number"},
    {NULL}  /* Sentinel */
};

static PyObject *
Noddy_name(Noddy* self)
{
    if (self->first == NULL) {
        PyErr_SetString(PyExc_AttributeError, "first");
        return NULL;
    }

    if (self->last == NULL) {
        PyErr_SetString(PyExc_AttributeError, "last");
        return NULL;
    }

    return PyUnicode_FromFormat("%S %S", self->first, self->last);
}

static PyMethodDef Noddy_methods[] = {
    {"name", (PyCFunction)Noddy_name, METH_NOARGS,
     "Return the name, combining the first and last name"
    },
    {NULL}  /* Sentinel */
};

static PyTypeObject NoddyType = {
    PyVarObject_HEAD_INIT(NULL, 0)
    "noddy.Noddy",             /* tp_name */
    sizeof(Noddy),             /* tp_basicsize */
    0,                         /* tp_itemsize */
    (destructor)Noddy_dealloc, /* tp_dealloc */
    0,                         /* tp_print */
    0,                         /* tp_getattr */
    0,                         /* tp_setattr */
    0,                         /* tp_reserved */
    0,                         /* tp_repr */
    0,                         /* tp_as_number */
    0,                         /* tp_as_sequence */
    0,                         /* tp_as_mapping */
    0,                         /* tp_hash  */
    0,                         /* tp_call */
    0,                         /* tp_str */
    0,                         /* tp_getattro */
    0,                         /* tp_setattro */
    0,                         /* tp_as_buffer */
    Py_TPFLAGS_DEFAULT |
        Py_TPFLAGS_BASETYPE |
        Py_TPFLAGS_HAVE_GC,    /* tp_flags */
    "Noddy objects",           /* tp_doc */
    (traverseproc)Noddy_traverse,   /* tp_traverse */
    (inquiry)Noddy_clear,           /* tp_clear */
    0,                         /* tp_richcompare */
    0,                         /* tp_weaklistoffset */
    0,                         /* tp_iter */
    0,                         /* tp_iternext */
    Noddy_methods,             /* tp_methods */
    Noddy_members,             /* tp_members */
    0,                         /* tp_getset */
    0,                         /* tp_base */
    0,                         /* tp_dict */
    0,                         /* tp_descr_get */
    0,                         /* tp_descr_set */
    0,                         /* tp_dictoffset */
    (initproc)Noddy_init,      /* tp_init */
    0,                         /* tp_alloc */
    Noddy_new,                 /* tp_new */
};

static PyModuleDef noddy4module = {
    PyModuleDef_HEAD_INIT,
    "noddy4",
    "Example module that creates an extension type.",
    -1,
    NULL, NULL, NULL, NULL, NULL
};

PyMODINIT_FUNC
PyInit_noddy4(void)
{
    PyObject* m;

    if (PyType_Ready(&NoddyType) < 0)
        return NULL;

    m = PyModule_Create(&noddy4module);
    if (m == NULL)
        return NULL;

    Py_INCREF(&NoddyType);
    PyModule_AddObject(m, "Noddy", (PyObject *)&NoddyType);
    return m;
}

The traversal method provides access to subobjects that could participate in cycles:

static int
Noddy_traverse(Noddy *self, visitproc visit, void *arg)
{
    int vret;

    if (self->first) {
        vret = visit(self->first, arg);
        if (vret != 0)
            return vret;
    }
    if (self->last) {
        vret = visit(self->last, arg);
        if (vret != 0)
            return vret;
    }

    return 0;
}

For each subobject that can participate in cycles, we need to call the visit() function, which is passed to the traversal method. The visit() function takes as arguments the subobject and the extra argument arg passed to the traversal method. It returns an integer value that must be returned if it is non-zero.

Python provides a Py_VISIT() macro that automates calling visit functions. With Py_VISIT(), Noddy_traverse() can be simplified:

static int
Noddy_traverse(Noddy *self, visitproc visit, void *arg)
{
    Py_VISIT(self->first);
    Py_VISIT(self->last);
    return 0;
}

注釈

Note that the tp_traverse implementation must name its arguments exactly visit and arg in order to use Py_VISIT(). This is to encourage uniformity across these boring implementations.

We also need to provide a method for clearing any subobjects that can participate in cycles.

static int
Noddy_clear(Noddy *self)
{
    PyObject *tmp;

    tmp = self->first;
    self->first = NULL;
    Py_XDECREF(tmp);

    tmp = self->last;
    self->last = NULL;
    Py_XDECREF(tmp);

    return 0;
}

Notice the use of a temporary variable in Noddy_clear(). We use the temporary variable so that we can set each member to NULL before decrementing its reference count. We do this because, as was discussed earlier, if the reference count drops to zero, we might cause code to run that calls back into the object. In addition, because we now support garbage collection, we also have to worry about code being run that triggers garbage collection. If garbage collection is run, our tp_traverse handler could get called. We can't take a chance of having Noddy_traverse() called when a member's reference count has dropped to zero and its value hasn't been set to NULL.

Python provides a Py_CLEAR() that automates the careful decrementing of reference counts. With Py_CLEAR(), the Noddy_clear() function can be simplified:

static int
Noddy_clear(Noddy *self)
{
    Py_CLEAR(self->first);
    Py_CLEAR(self->last);
    return 0;
}

Note that Noddy_dealloc() may call arbitrary functions through __del__ method or weakref callback. It means circular GC can be triggered inside the function. Since GC assumes reference count is not zero, we need to untrack the object from GC by calling PyObject_GC_UnTrack() before clearing members. Here is reimplemented deallocator which uses PyObject_GC_UnTrack() and Noddy_clear().

static void
Noddy_dealloc(Noddy* self)
{
    PyObject_GC_UnTrack(self);
    Noddy_clear(self);
    Py_TYPE(self)->tp_free((PyObject*)self);
}

Finally, we add the Py_TPFLAGS_HAVE_GC flag to the class flags:

Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE | Py_TPFLAGS_HAVE_GC, /* tp_flags */

That's pretty much it. If we had written custom tp_alloc or tp_free slots, we'd need to modify them for cyclic-garbage collection. Most extensions will use the versions automatically provided.

2.1.4. Subclassing other types

It is possible to create new extension types that are derived from existing types. It is easiest to inherit from the built in types, since an extension can easily use the PyTypeObject it needs. It can be difficult to share these PyTypeObject structures between extension modules.

In this example we will create a Shoddy type that inherits from the built-in list type. The new type will be completely compatible with regular lists, but will have an additional increment() method that increases an internal counter.

>>> import shoddy
>>> s = shoddy.Shoddy(range(3))
>>> s.extend(s)
>>> print(len(s))
6
>>> print(s.increment())
1
>>> print(s.increment())
2

#include <Python.h>

typedef struct {
    PyListObject list;
    int state;
} Shoddy;


static PyObject *
Shoddy_increment(Shoddy *self, PyObject *unused)
{
    self->state++;
    return PyLong_FromLong(self->state);
}


static PyMethodDef Shoddy_methods[] = {
    {"increment", (PyCFunction)Shoddy_increment, METH_NOARGS,
     PyDoc_STR("increment state counter")},
    {NULL,	NULL},
};

static int
Shoddy_init(Shoddy *self, PyObject *args, PyObject *kwds)
{
    if (PyList_Type.tp_init((PyObject *)self, args, kwds) < 0)
        return -1;
    self->state = 0;
    return 0;
}


static PyTypeObject ShoddyType = {
    PyVarObject_HEAD_INIT(NULL, 0)
    "shoddy.Shoddy",         /* tp_name */
    sizeof(Shoddy),          /* tp_basicsize */
    0,                       /* tp_itemsize */
    0,                       /* tp_dealloc */
    0,                       /* tp_print */
    0,                       /* tp_getattr */
    0,                       /* tp_setattr */
    0,                       /* tp_reserved */
    0,                       /* tp_repr */
    0,                       /* tp_as_number */
    0,                       /* tp_as_sequence */
    0,                       /* tp_as_mapping */
    0,                       /* tp_hash */
    0,                       /* tp_call */
    0,                       /* tp_str */
    0,                       /* tp_getattro */
    0,                       /* tp_setattro */
    0,                       /* tp_as_buffer */
    Py_TPFLAGS_DEFAULT |
        Py_TPFLAGS_BASETYPE, /* tp_flags */
    0,                       /* tp_doc */
    0,                       /* tp_traverse */
    0,                       /* tp_clear */
    0,                       /* tp_richcompare */
    0,                       /* tp_weaklistoffset */
    0,                       /* tp_iter */
    0,                       /* tp_iternext */
    Shoddy_methods,          /* tp_methods */
    0,                       /* tp_members */
    0,                       /* tp_getset */
    0,                       /* tp_base */
    0,                       /* tp_dict */
    0,                       /* tp_descr_get */
    0,                       /* tp_descr_set */
    0,                       /* tp_dictoffset */
    (initproc)Shoddy_init,   /* tp_init */
    0,                       /* tp_alloc */
    0,                       /* tp_new */
};

static PyModuleDef shoddymodule = {
    PyModuleDef_HEAD_INIT,
    "shoddy",
    "Shoddy module",
    -1,
    NULL, NULL, NULL, NULL, NULL
};

PyMODINIT_FUNC
PyInit_shoddy(void)
{
    PyObject *m;

    ShoddyType.tp_base = &PyList_Type;
    if (PyType_Ready(&ShoddyType) < 0)
        return NULL;

    m = PyModule_Create(&shoddymodule);
    if (m == NULL)
        return NULL;

    Py_INCREF(&ShoddyType);
    PyModule_AddObject(m, "Shoddy", (PyObject *) &ShoddyType);
    return m;
}

As you can see, the source code closely resembles the Noddy examples in previous sections. We will break down the main differences between them.

typedef struct {
    PyListObject list;
    int state;
} Shoddy;

The primary difference for derived type objects is that the base type's object structure must be the first value. The base type will already include the PyObject_HEAD() at the beginning of its structure.

When a Python object is a Shoddy instance, its PyObject* pointer can be safely cast to both PyListObject* and Shoddy*.

static int
Shoddy_init(Shoddy *self, PyObject *args, PyObject *kwds)
{
    if (PyList_Type.tp_init((PyObject *)self, args, kwds) < 0)
       return -1;
    self->state = 0;
    return 0;
}

In the __init__ method for our type, we can see how to call through to the __init__ method of the base type.

This pattern is important when writing a type with custom new and dealloc methods. The new method should not actually create the memory for the object with tp_alloc, that will be handled by the base class when calling its tp_new.

When filling out the PyTypeObject() for the Shoddy type, you see a slot for tp_base(). Due to cross platform compiler issues, you can't fill that field directly with the PyList_Type(); it can be done later in the module's init() function.

PyMODINIT_FUNC
PyInit_shoddy(void)
{
    PyObject *m;

    ShoddyType.tp_base = &PyList_Type;
    if (PyType_Ready(&ShoddyType) < 0)
        return NULL;

    m = PyModule_Create(&shoddymodule);
    if (m == NULL)
        return NULL;

    Py_INCREF(&ShoddyType);
    PyModule_AddObject(m, "Shoddy", (PyObject *) &ShoddyType);
    return m;
}

Before calling PyType_Ready(), the type structure must have the tp_base slot filled in. When we are deriving a new type, it is not necessary to fill out the tp_alloc slot with PyType_GenericNew() -- the allocate function from the base type will be inherited.

After that, calling PyType_Ready() and adding the type object to the module is the same as with the basic Noddy examples.

2.2. Type Methods

この節ではさまざまな実装可能なタイプメソッドと、それらが何をするものであるかについて、ざっと説明します。

以下は PyTypeObject の定義です。デバッグビルドでしか使われないいくつかのメンバは省いてあります:

typedef struct _typeobject {
    PyObject_VAR_HEAD
    const char *tp_name; /* For printing, in format "<module>.<name>" */
    Py_ssize_t tp_basicsize, tp_itemsize; /* For allocation */

    /* Methods to implement standard operations */

    destructor tp_dealloc;
    printfunc tp_print;
    getattrfunc tp_getattr;
    setattrfunc tp_setattr;
    PyAsyncMethods *tp_as_async; /* formerly known as tp_compare (Python 2)
                                    or tp_reserved (Python 3) */
    reprfunc tp_repr;

    /* Method suites for standard classes */

    PyNumberMethods *tp_as_number;
    PySequenceMethods *tp_as_sequence;
    PyMappingMethods *tp_as_mapping;

    /* More standard operations (here for binary compatibility) */

    hashfunc tp_hash;
    ternaryfunc tp_call;
    reprfunc tp_str;
    getattrofunc tp_getattro;
    setattrofunc tp_setattro;

    /* Functions to access object as input/output buffer */
    PyBufferProcs *tp_as_buffer;

    /* Flags to define presence of optional/expanded features */
    unsigned long tp_flags;

    const char *tp_doc; /* Documentation string */

    /* call function for all accessible objects */
    traverseproc tp_traverse;

    /* delete references to contained objects */
    inquiry tp_clear;

    /* rich comparisons */
    richcmpfunc tp_richcompare;

    /* weak reference enabler */
    Py_ssize_t tp_weaklistoffset;

    /* Iterators */
    getiterfunc tp_iter;
    iternextfunc tp_iternext;

    /* Attribute descriptor and subclassing stuff */
    struct PyMethodDef *tp_methods;
    struct PyMemberDef *tp_members;
    struct PyGetSetDef *tp_getset;
    struct _typeobject *tp_base;
    PyObject *tp_dict;
    descrgetfunc tp_descr_get;
    descrsetfunc tp_descr_set;
    Py_ssize_t tp_dictoffset;
    initproc tp_init;
    allocfunc tp_alloc;
    newfunc tp_new;
    freefunc tp_free; /* Low-level free-memory routine */
    inquiry tp_is_gc; /* For PyObject_IS_GC */
    PyObject *tp_bases;
    PyObject *tp_mro; /* method resolution order */
    PyObject *tp_cache;
    PyObject *tp_subclasses;
    PyObject *tp_weaklist;
    destructor tp_del;

    /* Type attribute cache version tag. Added in version 2.6 */
    unsigned int tp_version_tag;

    destructor tp_finalize;

} PyTypeObject;

Now that's a lot of methods. Don't worry too much though - if you have a type you want to define, the chances are very good that you will only implement a handful of these.

As you probably expect by now, we're going to go over this and give more information about the various handlers. We won't go in the order they are defined in the structure, because there is a lot of historical baggage that impacts the ordering of the fields; be sure your type initialization keeps the fields in the right order! It's often easiest to find an example that includes all the fields you need (even if they're initialized to 0) and then change the values to suit your new type.

const char *tp_name; /* For printing */

The name of the type - as mentioned in the last section, this will appear in various places, almost entirely for diagnostic purposes. Try to choose something that will be helpful in such a situation!

Py_ssize_t tp_basicsize, tp_itemsize; /* For allocation */

These fields tell the runtime how much memory to allocate when new objects of this type are created. Python has some built-in support for variable length structures (think: strings, lists) which is where the tp_itemsize field comes in. This will be dealt with later.

const char *tp_doc;

ここには Python スクリプトリファレンス obj.__doc__ が doc string を返すときの文字列 (あるいはそのアドレス) を入れます。

Now we come to the basic type methods---the ones most extension types will implement.

2.2.1. ファイナライズとメモリ解放

destructor tp_dealloc;

型のインスタンスの参照カウントがゼロになり、Python インタプリタがそれを潰して再利用したくなると、この関数が呼ばれます。解放すべきメモリをその型が保持していたり、それ以外にも実行すべき後処理がある場合は、それらをここに入れられます。オブジェクトそれ自体もここで解放される必要があります。この関数の例は、以下のようなものです:

static void
newdatatype_dealloc(newdatatypeobject * obj)
{
    free(obj->obj_UnderlyingDatatypePtr);
    Py_TYPE(obj)->tp_free(obj);
}

メモリ解放関数でひとつ重要なのは、処理待ちの例外にいっさい手をつけないことです。なぜなら、解放用の関数は Python インタプリタがスタックを元の状態に戻すときに呼ばれることが多いからです。そして (通常の関数からの復帰でなく) 例外のためにスタックが巻き戻されるときは、すでに発生している例外からメモリ解放関数を守るものはありません。解放用の関数がおこなう動作が追加の Python のコードを実行してしまうと、それらは例外が発生していることを検知するかもしれません。これはインタプリタが誤解させるエラーを発生させることにつながります。これを防ぐ正しい方法は、安全でない操作を実行する前に処理待ちの例外を保存しておき、終わったらそれを元に戻すことです。これは PyErr_Fetch() および PyErr_Restore() 関数を使うことによって可能になります:

static void
my_dealloc(PyObject *obj)
{
    MyObject *self = (MyObject *) obj;
    PyObject *cbresult;

    if (self->my_callback != NULL) {
        PyObject *err_type, *err_value, *err_traceback;

        /* This saves the current exception state */
        PyErr_Fetch(&err_type, &err_value, &err_traceback);

        cbresult = PyObject_CallObject(self->my_callback, NULL);
        if (cbresult == NULL)
            PyErr_WriteUnraisable(self->my_callback);
        else
            Py_DECREF(cbresult);

        /* This restores the saved exception state */
        PyErr_Restore(err_type, err_value, err_traceback);

        Py_DECREF(self->my_callback);
    }
    Py_TYPE(obj)->tp_free((PyObject*)self);
}

注釈

メモリ解放関数の中で安全に行えることにはいくつか制限があります。 1つ目は、その型が (tp_traverse および tp_clear を使って) ガベージコレクションをサポートしている場合、 tp_dealloc が呼び出されるまでに、消去されファイナライズされてしまうオブジェクトのメンバーが有り得ることです。 2つ目は、 tp_dealloc の中ではオブジェクトは不安定な状態にあることです: つまり参照カウントが0であるということです。 (上の例にあるような) 複雑なオブジェクトや API の呼び出しでは、 tp_dealloc を再度呼び出し、二重解放からクラッシュすることになるかもしれません。

Python 3.4 からは、複雑なファイナライズのコードは tp_dealloc に置かず、代わりに新しく導入された tp_finalize という型メソッドを使うことが推奨されています。

参考

PEP 442 で新しいファイナライズの仕組みが説明されています。

2.2.2. オブジェクト表現

Python では、オブジェクトの文字列表現を生成するのに 2つのやり方があります: repr() 関数を使う方法と、 str() 関数を使う方法です。 (print() 関数は単に str() を呼び出します。) これらのハンドラはどちらも省略できます。

reprfunc tp_repr;
reprfunc tp_str;

tp_repr ハンドラは呼び出されたインスタンスの文字列表現を格納した文字列オブジェクトを返す必要があります。簡単な例は以下のようなものです:

static PyObject *
newdatatype_repr(newdatatypeobject * obj)
{
    return PyUnicode_FromFormat("Repr-ified_newdatatype{{size:\%d}}",
                                obj->obj_UnderlyingDatatypePtr->size);
}

tp_repr ハンドラが指定されていなければ、インタプリタはその型の tp_name とそのオブジェクトの一意な識別値をもちいて文字列表現を作成します。

tp_str ハンドラと str() の関係は、上の tp_repr ハンドラと repr() の関係に相当します。つまり、これは Python のコードがオブジェクトのインスタンスに対して str() を呼び出したときに呼ばれます。この関数の実装は tp_repr ハンドラのそれと非常に似ていますが、得られる文字列表現は人間が読むことを意図されています。 tp_str が指定されていない場合、かわりに tp_repr ハンドラが使われます。

以下は簡単な例です:

static PyObject *
newdatatype_str(newdatatypeobject * obj)
{
    return PyUnicode_FromFormat("Stringified_newdatatype{{size:\%d}}",
                                obj->obj_UnderlyingDatatypePtr->size);
}

2.2.3. 属性を管理する

属性をもつどのオブジェクトに対しても、その型は、それらオブジェクトの属性をどのように解決するか制御する関数を提供する必要があります。必要な関数としては、属性を (それが定義されていれば) 取り出すものと、もうひとつは属性に (それが許可されていれば) 値を設定するものです。属性を削除するのは特殊なケースで、この場合は新しい値としてハンドラに NULL が渡されます。

Python は 2つの属性ハンドラの組をサポートしています。属性をもつ型はどちらか一組を実装するだけでよく、それらの違いは一方の組が属性の名前を char* として受け取るのに対してもう一方の組は属性の名前を PyObject* として受け取る、というものです。それぞれの型はその実装にとって都合がよい方を使えます。

getattrfunc  tp_getattr;        /* char * version */
setattrfunc  tp_setattr;
/* ... */
getattrofunc tp_getattro;       /* PyObject * version */
setattrofunc tp_setattro;

オブジェクトの属性へのアクセスがつねに (すぐあとで説明する) 単純な操作だけならば、 PyObject* を使って属性を管理する関数として、総称的 (generic) な実装を使えます。特定の型に特化した属性ハンドラの必要性は Python 2.2 からほとんど完全になくなりました。しかし、多くの例はまだ、この新しく使えるようになった総称的なメカニズムを使うよう更新されてはいません。

2.2.3.1. 総称的な属性を管理する

ほとんどの型は 単純な 属性を使うだけです。では、どのような属性が単純だといえるのでしょうか? それが満たすべき条件はごくわずかです:

1. PyType_Ready() が呼ばれたとき、すでに属性の名前がわかっていること。
2. 属性を参照したり設定したりするときに、特別な記録のための処理が必要でなく、また参照したり設定した値に対してどんな操作も実行する必要がないこと。

これらの条件は、属性の値や、値が計算されるタイミング、または格納されたデータがどの程度妥当なものであるかといったことになんら制約を課すものではないことに注意してください。

PyType_Ready() が呼ばれると、これはそのタイプオブジェクトに参照されている 3つのテーブルを使って、そのタイプオブジェクトの辞書中にデスクリプタ(descriptor) を作成します。各デスクリプタは、インスタンスオブジェクトの属性に対するアクセスを制御します。それぞれのテーブルはなくてもかまいません。もしこれら 3つがすべて NULL だと、その型のインスタンスはその基底型から継承した属性だけを持つことになります。また、 tp_getattro および tp_setattro が NULL のままだった場合も、基底型にこれらの属性の操作がまかせられます。

テーブルはタイプオブジェクト中の 3つのメンバとして宣言されています:

struct PyMethodDef *tp_methods;
struct PyMemberDef *tp_members;
struct PyGetSetDef *tp_getset;

tp_methods が NULL でない場合、これは PyMethodDef 構造体への配列を指している必要があります。テーブル中の各エントリは、つぎのような構造体のインスタンスです:

typedef struct PyMethodDef {
    const char  *ml_name;       /* method name */
    PyCFunction  ml_meth;       /* implementation function */
    int          ml_flags;      /* flags */
    const char  *ml_doc;        /* docstring */
} PyMethodDef;

その型が提供する各メソッドについてひとつのエントリを定義する必要があります。基底型から継承してきたメソッドについてはエントリは必要ありません。これの最後には、配列の終わりを示すための見張り番 (sentinel) として追加のエントリがひとつ必要です。この場合、 ml_name メンバが sentinel として使われ、その値は NULL でなければなりません。

2番目のテーブルは、インスタンス中に格納されるデータと直接対応づけられた属性を定義するのに使います。いくつもの C の原始的な型がサポートされており、アクセスを読み出し専用にも読み書き可能にもできます。このテーブルで使われる構造体は次のように定義されています:

typedef struct PyMemberDef {
    char *name;
    int   type;
    int   offset;
    int   flags;
    char *doc;
} PyMemberDef;

このテーブルの各エントリに対してデスクリプタ(descriptor)が作成され、値をインスタンスの構造体から抽出しうる型に対してそれらが追加されます。 type メンバは structmember.h ヘッダで定義された型のコードをひとつ含んでいる必要があります。この値は Python における値と C における値をどのように変換しあうかを定めるものです。 flags メンバはこの属性がどのようにアクセスされるかを制御するフラグを格納するのに使われます。

以下のフラグ用定数は structmember.h で定義されており、これらはビットごとの OR を取って組み合わせられます。

定数	意味
READONLY	絶対に変更できない。
READ_RESTRICTED	制限モード (restricted mode) では参照できない。
WRITE_RESTRICTED	制限モード (restricted mode) では変更できない。
RESTRICTED	制限モード (restricted mode) では参照も変更もできない。

tp_members を使ったひとつの面白い利用法は、実行時に使われるデスクリプタを作成しておき、単にテーブル中にテキストを置いておくことによって、この方法で定義されたすべての属性に doc string を関連付けられるようにすることです。アプリケーションはこのイントロスペクション用 API を使って、クラスオブジェクトからデスクリプタを取り出し、その __doc__ 属性を使って doc string を得られます。

tp_methods テーブルと同じように、ここでも name メンバの値を NULL にした見張り用エントリが必要です。

2.2.3.2. 特定の型に特化した属性の管理

話を単純にするため、ここでは char* を使ったバージョンのみを示します。name パラメータの型はインターフェイスとして char* を使うか PyObject* を使うかの違いしかありません。この例では、上の総称的な例と同じことを効率的にやりますが、 Python 2.2 で追加された総称的な型のサポートを使わずにやります。これはハンドラの関数がどのようにして呼ばれるのかを説明します。これで、たとえその機能を拡張する必要があるとき、何をどうすればいいかわかるでしょう。

tp_getattr ハンドラはオブジェクトが属性への参照を要求するときに呼ばれます。これは、そのクラスの __getattr__() メソッドが呼ばれるであろう状況と同じ状況下で呼び出されます。

以下に例を示します。:

static PyObject *
newdatatype_getattr(newdatatypeobject *obj, char *name)
{
    if (strcmp(name, "data") == 0)
    {
        return PyLong_FromLong(obj->data);
    }

    PyErr_Format(PyExc_AttributeError,
                 "'%.50s' object has no attribute '%.400s'",
                 tp->tp_name, name);
    return NULL;
}

tp_setattr ハンドラは、クラスのインスタンスの __setattr__() または __delattr__() メソッドが呼ばれるであろう状況で呼び出されます。ある属性が削除されるとき、3番目のパラメータは NULL になります。以下の例はたんに例外を発生させるものですが、もし本当にこれと同じことをしたいなら、 tp_setattr ハンドラを NULL に設定すべきです。

static int
newdatatype_setattr(newdatatypeobject *obj, char *name, PyObject *v)
{
    (void)PyErr_Format(PyExc_RuntimeError, "Read-only attribute: \%s", name);
    return -1;
}

2.2.4. オブジェクトの比較

richcmpfunc tp_richcompare;

tp_richcompare ハンドラは比較処理が要求されたときに呼び出されます。 __lt__() のような 拡張比較メソッド に類似しており、 PyObject_RichCompare() と PyObject_RichCompareBool() からも呼び出されます。

この関数は 2 つの Python オブジェクトと演算子を引数に取り、演算子は Py_EQ, Py_NE, Py_LE, Py_GT, Py_LT, Py_GT のどれか 1 つです。関数は指定された演算子に従って 2 つのオブジェクトを比較し、比較に成功した場合の Py_True および Py_False か、比較処理が実装されておらず、もう一方のオブジェクトの比較処理を試行することを示す Py_NotImplemented か、例外が設定された場合の NULL のいづれかを返します。

これは内部ポインタのサイズが等しければ等しいと見なすデータ型のサンプル実装です:

static PyObject *
newdatatype_richcmp(PyObject *obj1, PyObject *obj2, int op)
{
    PyObject *result;
    int c, size1, size2;

    /* code to make sure that both arguments are of type
       newdatatype omitted */

    size1 = obj1->obj_UnderlyingDatatypePtr->size;
    size2 = obj2->obj_UnderlyingDatatypePtr->size;

    switch (op) {
    case Py_LT: c = size1 <  size2; break;
    case Py_LE: c = size1 <= size2; break;
    case Py_EQ: c = size1 == size2; break;
    case Py_NE: c = size1 != size2; break;
    case Py_GT: c = size1 >  size2; break;
    case Py_GE: c = size1 >= size2; break;
    }
    result = c ? Py_True : Py_False;
    Py_INCREF(result);
    return result;
 }

2.2.5. 抽象的なプロトコルのサポート

Python はいくつもの 抽象的な “プロトコル”をサポートしています。これらを使用する特定のインターフェイスについては 抽象オブジェクトレイヤ (abstract objects layer) で解説されています。

これら多数の抽象的なインターフェイスは、Python の実装が開発される初期の段階で定義されていました。とりわけ数値や辞書、そしてシーケンスなどのプロトコルは最初から Python の一部だったのです。それ以外のプロトコルはその後追加されました。型の実装にあるいくつかのハンドラルーチンに依存するようなプロトコルのために、古いプロトコルはハンドラの入ったオプションのブロックとして定義し、型オブジェクトから参照するようになりました。タイプオブジェクトの主部に追加のスロットをもつ新しいプロトコルについては、フラグ用のビットを立てることでそれらのスロットが存在しており、インタプリタがチェックすべきであることを指示できます。(このフラグ用のビットは、そのスロットの値が非 NULL であることを示しているわけではありません。フラグはスロットの存在を示すのに使えますが、そのスロットはまだ埋まっていないかもしれないのです。)

PyNumberMethods   *tp_as_number;
PySequenceMethods *tp_as_sequence;
PyMappingMethods  *tp_as_mapping;

お使いのオブジェクトを数値やシーケンス、あるいは辞書のようにふるまうようにしたいならば、それぞれに C の PyNumberMethods 構造体、 PySequenceMethods 構造体、または PyMappingMethods 構造体のアドレスを入れます。これらに適切な値を入れても入れなくてもかまいません。これらを使った例は Python の配布ソースにある Objects でみつけることができるでしょう。

hashfunc tp_hash;

This function, if you choose to provide it, should return a hash number for an instance of your data type. Here is a moderately pointless example:

static long
newdatatype_hash(newdatatypeobject *obj)
{
    long result;
    result = obj->obj_UnderlyingDatatypePtr->size;
    result = result * 3;
    return result;
}

ternaryfunc tp_call;

この関数は、その型のインスタンスが「関数として呼び出される」ときに呼ばれます。たとえばもし obj1 にそのインスタンスが入っていて、Python スクリプトで obj1('hello') を実行したとすると、 tp_call ハンドラが呼ばれます。

この関数は 3つの引数をとります:

1. arg1 is the instance of the data type which is the subject of the call. If the call is obj1('hello'), then arg1 is obj1.
2. arg2 is a tuple containing the arguments to the call. You can use PyArg_ParseTuple() to extract the arguments.
3. arg3 is a dictionary of keyword arguments that were passed. If this is non-NULL and you support keyword arguments, use PyArg_ParseTupleAndKeywords() to extract the arguments. If you do not want to support keyword arguments and this is non-NULL, raise a TypeError with a message saying that keyword arguments are not supported.

Here is a desultory example of the implementation of the call function.

/* Implement the call function.
 *    obj1 is the instance receiving the call.
 *    obj2 is a tuple containing the arguments to the call, in this
 *         case 3 strings.
 */
static PyObject *
newdatatype_call(newdatatypeobject *obj, PyObject *args, PyObject *other)
{
    PyObject *result;
    char *arg1;
    char *arg2;
    char *arg3;

    if (!PyArg_ParseTuple(args, "sss:call", &arg1, &arg2, &arg3)) {
        return NULL;
    }
    result = PyUnicode_FromFormat(
        "Returning -- value: [\%d] arg1: [\%s] arg2: [\%s] arg3: [\%s]\n",
        obj->obj_UnderlyingDatatypePtr->size,
        arg1, arg2, arg3);
    return result;
}

/* Iterators */
getiterfunc tp_iter;
iternextfunc tp_iternext;

These functions provide support for the iterator protocol. Any object which wishes to support iteration over its contents (which may be generated during iteration) must implement the tp_iter handler. Objects which are returned by a tp_iter handler must implement both the tp_iter and tp_iternext handlers. Both handlers take exactly one parameter, the instance for which they are being called, and return a new reference. In the case of an error, they should set an exception and return NULL.

For an object which represents an iterable collection, the tp_iter handler must return an iterator object. The iterator object is responsible for maintaining the state of the iteration. For collections which can support multiple iterators which do not interfere with each other (as lists and tuples do), a new iterator should be created and returned. Objects which can only be iterated over once (usually due to side effects of iteration) should implement this handler by returning a new reference to themselves, and should also implement the tp_iternext handler. File objects are an example of such an iterator.

Iterator objects should implement both handlers. The tp_iter handler should return a new reference to the iterator (this is the same as the tp_iter handler for objects which can only be iterated over destructively). The tp_iternext handler should return a new reference to the next object in the iteration if there is one. If the iteration has reached the end, it may return NULL without setting an exception or it may set StopIteration; avoiding the exception can yield slightly better performance. If an actual error occurs, it should set an exception and return NULL.

2.2.6. 弱参照(Weak Reference)のサポート

One of the goals of Python's weak-reference implementation is to allow any type to participate in the weak reference mechanism without incurring the overhead on those objects which do not benefit by weak referencing (such as numbers).

For an object to be weakly referencable, the extension must include a PyObject* field in the instance structure for the use of the weak reference mechanism; it must be initialized to NULL by the object's constructor. It must also set the tp_weaklistoffset field of the corresponding type object to the offset of the field. For example, the instance type is defined with the following structure:

typedef struct {
    PyObject_HEAD
    PyClassObject *in_class;       /* The class object */
    PyObject      *in_dict;        /* A dictionary */
    PyObject      *in_weakreflist; /* List of weak references */
} PyInstanceObject;

The statically-declared type object for instances is defined this way:

PyTypeObject PyInstance_Type = {
    PyVarObject_HEAD_INIT(&PyType_Type, 0)
    0,
    "module.instance",

    /* Lots of stuff omitted for brevity... */

    Py_TPFLAGS_DEFAULT,                         /* tp_flags */
    0,                                          /* tp_doc */
    0,                                          /* tp_traverse */
    0,                                          /* tp_clear */
    0,                                          /* tp_richcompare */
    offsetof(PyInstanceObject, in_weakreflist), /* tp_weaklistoffset */
};

The type constructor is responsible for initializing the weak reference list to NULL:

static PyObject *
instance_new() {
    /* Other initialization stuff omitted for brevity */

    self->in_weakreflist = NULL;

    return (PyObject *) self;
}

The only further addition is that the destructor needs to call the weak reference manager to clear any weak references. This is only required if the weak reference list is non-NULL:

static void
instance_dealloc(PyInstanceObject *inst)
{
    /* Allocate temporaries if needed, but do not begin
       destruction just yet.
     */

    if (inst->in_weakreflist != NULL)
        PyObject_ClearWeakRefs((PyObject *) inst);

    /* Proceed with object destruction normally. */
}

2.2.7. その他いろいろ

Remember that you can omit most of these functions, in which case you provide 0 as a value. There are type definitions for each of the functions you must provide. They are in object.h in the Python include directory that comes with the source distribution of Python.

In order to learn how to implement any specific method for your new data type, do the following: Download and unpack the Python source distribution. Go to the Objects directory, then search the C source files for tp_ plus the function you want (for example, tp_richcompare). You will find examples of the function you want to implement.

When you need to verify that an object is an instance of the type you are implementing, use the PyObject_TypeCheck() function. A sample of its use might be something like the following:

if (! PyObject_TypeCheck(some_object, &MyType)) {
    PyErr_SetString(PyExc_TypeError, "arg #1 not a mything");
    return NULL;
}

Footnotes

[1]	This is true when we know that the object is a basic type, like a string or a float.

[2]	We relied on this in the tp_dealloc handler in this example, because our type doesn't support garbage collection. Even if a type supports garbage collection, there are calls that can be made to "untrack" the object from garbage collection, however, these calls are advanced and not covered here.

[3]

We now know that the first and last members are strings, so perhaps we could be less careful about decrementing their reference counts, however, we accept instances of string subclasses. Even though deallocating normal strings won't call back into our objects, we can't guarantee that deallocating an instance of a string subclass won't call back into our objects.

[4]	Even in the third version, we aren't guaranteed to avoid cycles. Instances of string subclasses are allowed and string subclasses could allow cycles even if normal strings don't.