Common Object Structures

There are a large number of structures which are used in the definition of object types for Python. This section describes these structures and how they are used.

Base object types and macros

All Python objects ultimately share a small number of fields at the beginning of the object's representation in memory. These are represented by the PyObject and PyVarObject types, which are defined, in turn, by the expansions of some macros also used, whether directly or indirectly, in the definition of all other Python objects. Additional macros can be found under reference counting <countingrefs>.

All object types are extensions of this type. This is a type which contains the information Python needs to treat a pointer to an object as an object. In a normal "release" build, it contains only the object's reference count and a pointer to the corresponding type object. Nothing is actually declared to be a PyObject, but every pointer to a Python object can be cast to a PyObject*.

The members must not be accessed directly; instead use macros such as Py_REFCNT and Py_TYPE.

In the Stable ABI <stable-abi> for Free-Threaded Builds (abi3t), this struct is opaque; its size and layout may change between Python versions. In Stable ABI for non-free-threaded builds (abi3), the !ob_refcnt and !ob_type fields are available, but using them directly is discouraged.

The object's reference count, as returned by Py_REFCNT. Do not use this field directly; instead use functions and macros such as !Py_REFCNT, Py_INCREF and Py_DecRef.

The field type may be different from Py_ssize_t, depending on build configuration and platform.

The object's type. Do not use this field directly; use Py_TYPE and Py_SET_TYPE instead.

A per-object lock <per-object-locks>, present only in the free-threaded <free threading> build (when Py_GIL_DISABLED is defined).

This field is reserved for use by the critical section API (Py_BEGIN_CRITICAL_SECTION / Py_END_CRITICAL_SECTION). Do not lock it directly with PyMutex_Lock; doing so can cause deadlocks. If you need your own lock, add a separate PyMutex field to your object struct.

3.13

An extension of PyObject that adds the ~PyVarObject.ob_size field. This is intended for objects that have some notion of length.

As with !PyObject, the members must not be accessed directly; instead use macros such as Py_SIZE, Py_REFCNT and Py_TYPE.

In the Stable ABI <stable-abi> for Free-Threaded Builds (abi3t), this struct is opaque; its size and layout may change between Python versions. In Stable ABI for non-free-threaded builds (abi3), the !ob_base and !ob_size fields are available, but using them directly is discouraged.

Common object header. Typically, this field is not accessed directly; instead !PyVarObject can be cast to PyObject.

A size field, whose contents should be considered an object's internal implementation detail.

Do not use this field directly; use Py_SIZE instead.

Object creation functions such as PyObject_NewVar will generally set this field to the requested size (number of items). After creation, arbitrary values can be stored in !ob_size using Py_SET_SIZE.

To get an object's publicly exposed length, as returned by the Python function len, use PyObject_Length instead.

This is a macro used when declaring new types which represent objects without a varying length. The PyObject_HEAD macro expands to:

PyObject ob_base;

See documentation of PyObject above.

This is a macro used when declaring new types which represent objects with a length that varies from instance to instance. The PyObject_VAR_HEAD macro expands to:

PyVarObject ob_base;

See documentation of PyVarObject above.

The base class of all other objects, the same as object in Python.

Test if the x object is the y object, the same as x is y in Python.

3.10

Test if an object is the None singleton, the same as x is None in Python.

3.10

Test if an object is the True singleton, the same as x is True in Python.

3.10

Test if an object is the False singleton, the same as x is False in Python.

3.10

Get the type of the Python object o.

The returned reference is borrowed <borrowed reference> from o. Do not release it with Py_DECREF or similar.

3.11 Py_TYPE() is changed to an inline static function. The parameter type is no longer const PyObject*.

Return non-zero if the object o type is type. Return zero otherwise. Equivalent to: Py_TYPE(o) == type.

3.9

Set the type of object o to type, without any checking or reference counting.

This is a very low-level operation. Consider instead setting the Python attribute ~object.__class__ using PyObject_SetAttrString or similar.

Note that assigning an incompatible type can lead to undefined behavior.

If type is a heap type <heap-types>, the caller must create a new reference to it. Similarly, if the old type of o is a heap type, the caller must release a reference to that type.

3.9

Get the ~PyVarObject.ob_size field of o.

3.11 Py_SIZE() is changed to an inline static function. The parameter type is no longer const PyVarObject*.

Set the ~PyVarObject.ob_size field of o to size.

3.9

This is a macro which expands to initialization values for a new PyObject type. This macro expands to:

_PyObject_EXTRA_INIT
1, type,

This is a macro which expands to initialization values for a new PyVarObject type, including the ~PyVarObject.ob_size field. This macro expands to:

_PyObject_EXTRA_INIT
1, type, size,

Implementing functions and methods

Type of the functions used to implement most Python callables in C. Functions of this type take two PyObject* parameters and return one such value. If the return value is NULL, an exception shall have been set. If not NULL, the return value is interpreted as the return value of the function as exposed in Python. The function must return a new reference.

The function signature is:

PyObject *PyCFunction(PyObject *self,
                      PyObject *args);

Type of the functions used to implement Python callables in C with signature METH_VARARGS | METH_KEYWORDS <METH_VARARGS-METH_KEYWORDS>. The function signature is:

PyObject *PyCFunctionWithKeywords(PyObject *self,
                                  PyObject *args,
                                  PyObject *kwargs);

Type of the functions used to implement Python callables in C with signature METH_FASTCALL. The function signature is:

PyObject *PyCFunctionFast(PyObject *self,
                          PyObject *const *args,
                          Py_ssize_t nargs);

Type of the functions used to implement Python callables in C with signature METH_FASTCALL | METH_KEYWORDS <METH_FASTCALL-METH_KEYWORDS>. The function signature is:

PyObject *PyCFunctionFastWithKeywords(PyObject *self,
                                      PyObject *const *args,
                                      Py_ssize_t nargs,
                                      PyObject *kwnames);

Type of the functions used to implement Python callables in C with signature METH_METHOD | METH_FASTCALL | METH_KEYWORDS <METH_METHOD-METH_FASTCALL-METH_KEYWORDS>. The function signature is:

PyObject *PyCMethod(PyObject *self,
                    PyTypeObject *defining_class,
                    PyObject *const *args,
                    Py_ssize_t nargs,
                    PyObject *kwnames)

3.9

Structure used to describe a method of an extension type. This structure has four fields:

Name of the method.

A NULL ml_name marks the end of a !PyMethodDef array.

Pointer to the C implementation.

Flags bits indicating how the call should be constructed.

Points to the contents of the docstring.

The ~PyMethodDef.ml_meth is a C function pointer. The functions may be of different types, but they always return PyObject*. If the function is not of the PyCFunction, the compiler will require a cast in the method table. Even though PyCFunction defines the first parameter as PyObject*, it is common that the method implementation uses the specific C type of the self object.

The ~PyMethodDef.ml_flags field is a bitfield which can include the following flags. The individual flags indicate either a calling convention or a binding convention.

There are these calling conventions:

This is the typical calling convention, where the methods have the type PyCFunction. The function expects two PyObject* values. The first one is the self object for methods; for module functions, it is the module object. The second parameter (often called args) is a tuple object representing all arguments. This parameter is typically processed using PyArg_ParseTuple or PyArg_UnpackTuple.

Can only be used in certain combinations with other flags: METH_VARARGS | METH_KEYWORDS <METH_VARARGS-METH_KEYWORDS>, METH_FASTCALL | METH_KEYWORDS <METH_FASTCALL-METH_KEYWORDS> and METH_METHOD | METH_FASTCALL | METH_KEYWORDS <METH_METHOD-METH_FASTCALL-METH_KEYWORDS>.

Fast calling convention supporting only positional arguments. The methods have the type PyCFunctionFast. The first parameter is self, the second parameter is a C array of PyObject* values indicating the arguments and the third parameter is the number of arguments (the length of the array).

3.7

3.10

METH_FASTCALL is now part of the stable ABI <stable-abi>.

Can only be used in the combination with other flags: METH_METHOD | METH_FASTCALL | METH_KEYWORDS <METH_METHOD-METH_FASTCALL-METH_KEYWORDS>.

Methods without parameters don't need to check whether arguments are given if they are listed with the METH_NOARGS flag. They need to be of type PyCFunction. The first parameter is typically named self and will hold a reference to the module or object instance. In all cases the second parameter will be NULL.

The function must have 2 parameters. Since the second parameter is unused, Py_UNUSED can be used to prevent a compiler warning.

Methods with a single object argument can be listed with the METH_O flag, instead of invoking PyArg_ParseTuple with a "O" argument. They have the type PyCFunction, with the self parameter, and a PyObject* parameter representing the single argument.

These two constants are not used to indicate the calling convention but the binding when used with methods of classes. These may not be used for functions defined for modules. At most one of these flags may be set for any given method.

pair: built-in function; classmethod

The method will be passed the type object as the first parameter rather than an instance of the type. This is used to create class methods, similar to what is created when using the classmethod built-in decorator.

pair: built-in function; staticmethod

The method will be passed NULL as the first parameter rather than an instance of the type. This is used to create static methods, similar to what is created when using the staticmethod built-in decorator.

One other constant controls whether a method is loaded in place of another definition with the same method name.

The method will be loaded in place of existing definitions. Without METH_COEXIST, the default is to skip repeated definitions. Since slot wrappers are loaded before the method table, the existence of a sq_contains slot, for example, would generate a wrapped method named ~object.__contains__ and preclude the loading of a corresponding PyCFunction with the same name. With the flag defined, the PyCFunction will be loaded in place of the wrapper object and will co-exist with the slot. This is helpful because calls to PyCFunctions are optimized more than wrapper object calls.

The type object corresponding to Python C method objects. This is available as types.BuiltinMethodType in the Python layer.

Return true if op is an instance of the PyCMethod_Type type or a subtype of it. This function always succeeds.

This is the same as PyCMethod_Check, but does not account for subtypes.

Turn ml into a Python callable object. The caller must ensure that ml outlives the callable. Typically, ml is defined as a static variable.

The self parameter will be passed as the self argument to the C function in ml->ml_meth when invoked. self can be NULL.

The callable object's __module__ attribute can be set from the given module argument. module should be a Python string, which will be used as name of the module the function is defined in. If unavailable, it can be set to None or NULL.

function.__module__

The cls parameter will be passed as the defining_class argument to the C function. Must be set if METH_METHOD is set on ml->ml_flags.

3.9

The type object corresponding to Python C function objects. This is available as types.BuiltinFunctionType in the Python layer.

Return true if op is an instance of the PyCFunction_Type type or a subtype of it. This function always succeeds.

This is the same as PyCFunction_Check, but does not account for subtypes.

Equivalent to PyCMethod_New(ml, self, module, NULL).

Equivalent to PyCMethod_New(ml, self, NULL, NULL).

Get the function's flags on func as they were passed to ~PyMethodDef.ml_flags.

If func is not a C function object, this fails with an exception. func must not be NULL.

This function returns the function's flags on success, and -1 with an exception set on failure.

This is the same as PyCFunction_GetFlags, but without error or type checking.

Get the function pointer on func as it was passed to ~PyMethodDef.ml_meth.

If func is not a C function object, this fails with an exception. func must not be NULL.

This function returns the function pointer on success, and NULL with an exception set on failure.

This is the same as PyCFunction_GetFunction, but without error or type checking.

Get the "self" object on func. This is the object that would be passed to the first argument of a PyCFunction. For C function objects created through a PyMethodDef on a PyModuleDef, this is the resulting module object.

If func is not a C function object, this fails with an exception. func must not be NULL.

This function returns a borrowed reference to the "self" object on success, and NULL with an exception set on failure.

This is the same as PyCFunction_GetSelf, but without error or type checking.

Return a borrowed reference to the defining class of func if its ~PyMethodDef.ml_flags includes METH_METHOD. Otherwise, return NULL without setting an exception.

func must be a C function object and must not be NULL. This function does not perform error or type checking.

3.9

Accessing attributes of extension types

Structure which describes an attribute of a type which corresponds to a C struct member. When defining a class, put a NULL-terminated array of these structures in the ~PyTypeObject.tp_members slot.

Its fields are, in order:

Name of the member. A NULL value marks the end of a PyMemberDef[] array.

The string should be static, no copy is made of it.

The type of the member in the C struct. See PyMemberDef-types for the possible values.

The offset in bytes that the member is located on the type’s object struct.

Zero or more of the PyMemberDef-flags, combined using bitwise OR.

The docstring, or NULL. The string should be static, no copy is made of it. Typically, it is defined using PyDoc_STR.

By default (when ~PyMemberDef.flags is 0), members allow both read and write access. Use the Py_READONLY flag for read-only access. Certain types, like Py_T_STRING, imply Py_READONLY. Only Py_T_OBJECT_EX (and legacy T_OBJECT) members can be deleted.

For heap-allocated types (created using PyType_FromSpec or similar), PyMemberDef may contain a definition for the special member "__vectorcalloffset__", corresponding to ~PyTypeObject.tp_vectorcall_offset in type objects. This member must be defined with Py_T_PYSSIZET, and either Py_READONLY or Py_READONLY | Py_RELATIVE_OFFSET. For example:

static PyMemberDef spam_type_members[] = {
    {"__vectorcalloffset__", Py_T_PYSSIZET,
     offsetof(Spam_object, vectorcall), Py_READONLY},
    {NULL}  /* Sentinel */
};

(You may need to #include <stddef.h> for !offsetof.)

The legacy offsets ~PyTypeObject.tp_dictoffset and ~PyTypeObject.tp_weaklistoffset can be defined similarly using "__dictoffset__" and "__weaklistoffset__" members, but extensions are strongly encouraged to use Py_TPFLAGS_MANAGED_DICT and Py_TPFLAGS_MANAGED_WEAKREF instead.

3.12

PyMemberDef is always available. Previously, it required including "structmember.h".

3.14

Py_RELATIVE_OFFSET is now allowed for "__vectorcalloffset__", "__dictoffset__" and "__weaklistoffset__".

Get an attribute belonging to the object at address obj_addr. The attribute is described by PyMemberDef m. Returns NULL on error.

3.12

PyMember_GetOne is always available. Previously, it required including "structmember.h".

Set an attribute belonging to the object at address obj_addr to object o. The attribute to set is described by PyMemberDef m. Returns 0 if successful and a negative value on failure.

3.12

PyMember_SetOne is always available. Previously, it required including "structmember.h".

Member flags

The following flags can be used with PyMemberDef.flags:

Not writable.

Emit an object.__getattr__ audit event <audit-events> before reading.

Indicates that the ~PyMemberDef.offset of this PyMemberDef entry indicates an offset from the subclass-specific data, rather than from PyObject.

Can only be used as part of the Py_tp_members slot <PyType_Slot> when creating a class using negative ~PyType_Spec.basicsize. It is mandatory in that case. When setting ~PyTypeObject.tp_members from the slot during class creation, Python clears the flag and sets PyMemberDef.offset to the offset from the PyObject struct.

Soft-deprecated alias of Py_AUDIT_READ (note capitalization of the y). Only available with #include "structmember.h".

3.12 Use Py_AUDIT_READ instead.

A member flag that has no effect, retained solely for backwards compatibility. This is only available with #include "structmember.h".

next Omit this flag instead.

Member types

PyMemberDef.type can be one of the following macros corresponding to various C types. When the member is accessed in Python, it will be converted to the equivalent Python type. When it is set from Python, it will be converted back to the C type. If that is not possible, an exception such as TypeError or ValueError is raised.

Unless marked (D), attributes defined this way cannot be deleted using e.g. del or delattr.

(*): Zero-terminated, UTF8-encoded C string. With !Py_T_STRING the C representation is a pointer; with !Py_T_STRING_INPLACE the string is stored directly in the structure.

(**): String of length 1. Only ASCII is accepted.

(RO): Implies Py_READONLY.

(D): Can be deleted, in which case the pointer is set to NULL. Reading a NULL pointer raises AttributeError.

Defining Getters and Setters

Structure to define property-like access for a type. See also description of the PyTypeObject.tp_getset slot.

attribute name

C function to get the attribute.

Optional C function to set or delete the attribute. If NULL, the attribute is read-only.

optional docstring

Optional user data pointer, providing additional data for getter and setter.

The get function takes one PyObject* parameter (the instance) and a user data pointer (the associated closure):

It should return a new reference on success or NULL with a set exception on failure.

set functions take two PyObject* parameters (the instance and the value to be set) and a user data pointer (the associated closure):

In case the attribute should be deleted the second parameter is NULL. Should return 0 on success or -1 with a set exception on failure.