Doc/api/newtypes.tex

\chapter{Object Implementation Support \label{newTypes}}


This chapter describes the functions, types, and macros used when
defining new object types.


\section{Allocating Objects on the Heap
         \label{allocating-objects}}

\begin{cfuncdesc}{PyObject*}{_PyObject_New}{PyTypeObject *type}
\end{cfuncdesc}

\begin{cfuncdesc}{PyVarObject*}{_PyObject_NewVar}{PyTypeObject *type, int size}
\end{cfuncdesc}

\begin{cfuncdesc}{void}{_PyObject_Del}{PyObject *op}
\end{cfuncdesc}

\begin{cfuncdesc}{PyObject*}{PyObject_Init}{PyObject *op,
					    PyTypeObject *type}
  Initialize a newly-allocated object \var{op} with its type and
  initial reference.  Returns the initialized object.  If \var{type}
  indicates that the object participates in the cyclic garbage
  detector, it it added to the detector's set of observed objects.
  Other fields of the object are not affected.
\end{cfuncdesc}

\begin{cfuncdesc}{PyVarObject*}{PyObject_InitVar}{PyVarObject *op,
						  PyTypeObject *type, int size}
  This does everything \cfunction{PyObject_Init()} does, and also
  initializes the length information for a variable-size object.
\end{cfuncdesc}

\begin{cfuncdesc}{\var{TYPE}*}{PyObject_New}{TYPE, PyTypeObject *type}
  Allocate a new Python object using the C structure type \var{TYPE}
  and the Python type object \var{type}.  Fields not defined by the
  Python object header are not initialized; the object's reference
  count will be one.  The size of the memory
  allocation is determined from the \member{tp_basicsize} field of the
  type object.
\end{cfuncdesc}

\begin{cfuncdesc}{\var{TYPE}*}{PyObject_NewVar}{TYPE, PyTypeObject *type,
                                                int size}
  Allocate a new Python object using the C structure type \var{TYPE}
  and the Python type object \var{type}.  Fields not defined by the
  Python object header are not initialized.  The allocated memory
  allows for the \var{TYPE} structure plus \var{size} fields of the
  size given by the \member{tp_itemsize} field of \var{type}.  This is
  useful for implementing objects like tuples, which are able to
  determine their size at construction time.  Embedding the array of
  fields into the same allocation decreases the number of allocations,
  improving the memory management efficiency.
\end{cfuncdesc}

\begin{cfuncdesc}{void}{PyObject_Del}{PyObject *op}
  Releases memory allocated to an object using
  \cfunction{PyObject_New()} or \cfunction{PyObject_NewVar()}.  This
  is normally called from the \member{tp_dealloc} handler specified in
  the object's type.  The fields of the object should not be accessed
  after this call as the memory is no longer a valid Python object.
\end{cfuncdesc}

\begin{cfuncdesc}{\var{TYPE}*}{PyObject_NEW}{TYPE, PyTypeObject *type}
  Macro version of \cfunction{PyObject_New()}, to gain performance at
  the expense of safety.  This does not check \var{type} for a \NULL{}
  value.
\end{cfuncdesc}

\begin{cfuncdesc}{\var{TYPE}*}{PyObject_NEW_VAR}{TYPE, PyTypeObject *type,
                                                int size}
  Macro version of \cfunction{PyObject_NewVar()}, to gain performance
  at the expense of safety.  This does not check \var{type} for a
  \NULL{} value.
\end{cfuncdesc}

\begin{cfuncdesc}{void}{PyObject_DEL}{PyObject *op}
  Macro version of \cfunction{PyObject_Del()}.
\end{cfuncdesc}

\begin{cfuncdesc}{PyObject*}{Py_InitModule}{char *name,
                                            PyMethodDef *methods}
  Create a new module object based on a name and table of functions,
  returning the new module object.
\end{cfuncdesc}

\begin{cfuncdesc}{PyObject*}{Py_InitModule3}{char *name,
                                             PyMethodDef *methods,
                                             char *doc}
  Create a new module object based on a name and table of functions,
  returning the new module object.  If \var{doc} is non-\NULL, it will
  be used to define the docstring for the module.
\end{cfuncdesc}

\begin{cfuncdesc}{PyObject*}{Py_InitModule4}{char *name,
                                             PyMethodDef *methods,
                                             char *doc, PyObject *self,
                                             int apiver}
  Create a new module object based on a name and table of functions,
  returning the new module object.  If \var{doc} is non-\NULL, it will
  be used to define the docstring for the module.  If \var{self} is
  non-\NULL, it will passed to the functions of the module as their
  (otherwise \NULL) first parameter.  (This was added as an
  experimental feature, and there are no known uses in the current
  version of Python.)  For \var{apiver}, the only value which should
  be passed is defined by the constant \constant{PYTHON_API_VERSION}.

  \note{Most uses of this function should probably be using
  the \cfunction{Py_InitModule3()} instead; only use this if you are
  sure you need it.}
\end{cfuncdesc}

DL_IMPORT

\begin{cvardesc}{PyObject}{_Py_NoneStruct}
  Object which is visible in Python as \code{None}.  This should only
  be accessed using the \code{Py_None} macro, which evaluates to a
  pointer to this object.
\end{cvardesc}


\section{Common Object Structures \label{common-structs}}

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.

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 \ctype{PyObject} and \ctype{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.

\begin{ctypedesc}{PyObject}
  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 objects reference count and a pointer to the corresponding
  type object.  It corresponds to the fields defined by the
  expansion of the \code{PyObject_VAR_HEAD} macro.
\end{ctypedesc}

\begin{ctypedesc}{PyVarObject}
  This is an extension of \ctype{PyObject} that adds the
  \member{ob_size} field.  This is only used for objects that have
  some notion of \emph{length}.  This type does not often appear in
  the Python/C API.  It corresponds to the fields defined by the
  expansion of the \code{PyObject_VAR_HEAD} macro.
\end{ctypedesc}

These macros are used in the definition of \ctype{PyObject} and
\ctype{PyVarObject}:

\begin{csimplemacrodesc}{PyObject_HEAD}
  This is a macro which expands to the declarations of the fields of
  the \ctype{PyObject} type; it is used when declaring new types which
  represent objects without a varying length.  The specific fields it
  expands to depends on the definition of
  \csimplemacro{Py_TRACE_REFS}.  By default, that macro is not
  defined, and \csimplemacro{PyObject_HEAD} expands to:
  \begin{verbatim}
    int ob_refcnt;
    PyTypeObject *ob_type;
  \end{verbatim}
  When \csimplemacro{Py_TRACE_REFS} is defined, it expands to:
  \begin{verbatim}
    PyObject *_ob_next, *_ob_prev;
    int ob_refcnt;
    PyTypeObject *ob_type;
  \end{verbatim}
\end{csimplemacrodesc}

\begin{csimplemacrodesc}{PyObject_VAR_HEAD}
  This is a macro which expands to the declarations of the fields of
  the \ctype{PyVarObject} type; it is used when declaring new types which
  represent objects with a length that varies from instance to
  instance.  This macro always expands to:
  \begin{verbatim}
    PyObject_HEAD
    int ob_size;
  \end{verbatim}
  Note that \csimplemacro{PyObject_HEAD} is part of the expansion, and
  that it's own expansion varies depending on the definition of
  \csimplemacro{Py_TRACE_REFS}.
\end{csimplemacrodesc}

PyObject_HEAD_INIT

\begin{ctypedesc}{PyCFunction}
  Type of the functions used to implement most Python callables in C.
  Functions of this type take two \ctype{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.
\end{ctypedesc}

\begin{ctypedesc}{PyMethodDef}
  Structure used to describe a method of an extension type.  This
  structure has four fields:

  \begin{tableiii}{l|l|l}{member}{Field}{C Type}{Meaning}
    \lineiii{ml_name}{char *}{name of the method}
    \lineiii{ml_meth}{PyCFunction}{pointer to the C implementation}
    \lineiii{ml_flags}{int}{flag bits indicating how the call should be
                            constructed}
    \lineiii{ml_doc}{char *}{points to the contents of the docstring}
  \end{tableiii}
\end{ctypedesc}

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

The \member{ml_flags} field is a bitfield which can include the
following flags.  The individual flags indicate either a calling
convention or a binding convention.  Of the calling convention flags,
only \constant{METH_VARARGS} and \constant{METH_KEYWORDS} can be
combined.  Any of the calling convention flags can be combined with a
binding flag.

\begin{datadesc}{METH_VARARGS}
  This is the typical calling convention, where the methods have the
  type \ctype{PyMethodDef}. The function expects two
  \ctype{PyObject*} values.  The first one is the \var{self} object for
  methods; for module functions, it has the value given to
  \cfunction{Py_InitModule4()} (or \NULL{} if
  \cfunction{Py_InitModule()} was used).  The second parameter
  (often called \var{args}) is a tuple object representing all
  arguments. This parameter is typically processed using
  \cfunction{PyArg_ParseTuple()}.
\end{datadesc}

\begin{datadesc}{METH_KEYWORDS}
  Methods with these flags must be of type
  \ctype{PyCFunctionWithKeywords}.  The function expects three
  parameters: \var{self}, \var{args}, and a dictionary of all the
  keyword arguments.  The flag is typically combined with
  \constant{METH_VARARGS}, and the parameters are typically processed
  using \cfunction{PyArg_ParseTupleAndKeywords()}.
\end{datadesc}

\begin{datadesc}{METH_NOARGS}
  Methods without parameters don't need to check whether arguments are
  given if they are listed with the \constant{METH_NOARGS} flag.  They
  need to be of type \ctype{PyNoArgsFunction}: they expect a single
  single \ctype{PyObject*} as a parameter.  When used with object
  methods, this parameter is typically named \code{self} and will hold
  a reference to the object instance.
\end{datadesc}

\begin{datadesc}{METH_O}
  Methods with a single object argument can be listed with the
  \constant{METH_O} flag, instead of invoking
  \cfunction{PyArg_ParseTuple()} with a \code{"O"} argument. They have
  the type \ctype{PyCFunction}, with the \var{self} parameter, and a
  \ctype{PyObject*} parameter representing the single argument.
\end{datadesc}

\begin{datadesc}{METH_OLDARGS}
  This calling convention is deprecated.  The method must be of type
  \ctype{PyCFunction}.  The second argument is \NULL{} if no arguments
  are given, a single object if exactly one argument is given, and a
  tuple of objects if more than one argument is given.  There is no
  way for a function using this convention to distinguish between a
  call with multiple arguments and a call with a tuple as the only
  argument.
\end{datadesc}

These two constants are not used to indicate the calling convention
but the binding when use 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.

\begin{datadesc}{METH_CLASS}
  The method will be passed the type object as the first parameter
  rather than an instance of the type.  This is used to create
  \emph{class methods}, similar to what is created when using the
  \function{classmethod()}\bifuncindex{classmethod} built-in
  function.
  \versionadded{2.3}
\end{datadesc}

\begin{datadesc}{METH_STATIC}
  The method will be passed \NULL{} as the first parameter rather than
  an instance of the type.  This is used to create \emph{static
  methods}, similar to what is created when using the
  \function{staticmethod()}\bifuncindex{staticmethod} built-in
  function.
  \versionadded{2.3}
\end{datadesc}


\begin{cfuncdesc}{PyObject*}{Py_FindMethod}{PyMethodDef table[],
                                            PyObject *ob, char *name}
  Return a bound method object for an extension type implemented in
  C.  This can be useful in the implementation of a
  \member{tp_getattro} or \member{tp_getattr} handler that does not
  use the \cfunction{PyObject_GenericGetAttr()} function.
\end{cfuncdesc}


\section{Type Objects \label{type-structs}}

Perhaps one of the most important structures of the Python object
system is the structure that defines a new type: the
\ctype{PyTypeObject} structure.  Type objects can be handled using any
of the \cfunction{PyObject_*()} or \cfunction{PyType_*()} functions,
but do not offer much that's interesting to most Python applications.
These objects are fundamental to how objects behave, so they are very
important to the interpreter itself and to any extension module that
implements new types.

Type objects are fairly large compared to most of the standard types.
The reason for the size is that each type object stores a large number
of values, mostly C function pointers, each of which implements a
small part of the type's functionality.  The fields of the type object
are examined in detail in this section.  The fields will be described
in the order in which they occur in the structure.

Typedefs:
unaryfunc, binaryfunc, ternaryfunc, inquiry, coercion, intargfunc,
intintargfunc, intobjargproc, intintobjargproc, objobjargproc,
destructor, freefunc, printfunc, getattrfunc, getattrofunc, setattrfunc,
setattrofunc, cmpfunc, reprfunc, hashfunc

The structure definition for \ctype{PyTypeObject} can be found in
\file{Include/object.h}.  For convenience of reference, this repeats
the definition found there:

\verbatiminput{typestruct.h}

The type object structure extends the \ctype{PyVarObject} structure,
though it does not actually need the the \member{ob_size} field.  The
inclusion of this field is a historical accident that must be
maintained to ensure binary compatibility between new versions of
Python and older compiled extensions.

\begin{cmemberdesc}{PyObject}{PyObject*}{_ob_next}
\cmemberline{PyObject}{PyObject*}{_ob_prev}
  These fields are only present when the macro \code{Py_TRACE_REFS} is
  defined.  Their initialization to \NULL{} is taken care of by the
  \code{PyObject_HEAD_INIT} macro.  For statically allocated objects,
  these fields always remain \NULL.  For dynamically allocated
  objects, these two fields are used to link the object into a
  doubly-linked list of \emph{all} live objects on the heap.  This
  could be used for various debugging purposes; currently the only use
  is to print the objects that are still alive at the end of a run
  when the environment variable \envvar{PYTHONDUMPREFS} is set.

  These fields are not inherited by subtypes.
\end{cmemberdesc}

\begin{cmemberdesc}{PyObject}{int}{ob_refcnt}
  This is the type object's reference count, initialized to \code{1}
  by the \code{PyObject_HEAD_INIT} macro.  Note that for statically
  allocated type objects, the type's instances (objects whose
  \member{ob_type} points back to the type) do \emph{not} count as
  references.  But for dynamically allocated type objects, the
  instances \emph{do} count as references.

  This field is not inherited by subtypes.
\end{cmemberdesc}

\begin{cmemberdesc}{PyObject}{PyTypeObject*}{ob_type}
  This is the type's type, in other words its metatype.  It is
  initialized by the argument to the \code{PyObject_HEAD_INIT} macro,
  and its value should normally be \code{\&PyType_Type}.  However, for
  dynamically loadable extension modules that must be usable on
  Windows (at least), the compiler complains that this is not a valid
  initializer.  Therefore, the convention is to pass \NULL{} to the
  \code{PyObject_HEAD_INIT} macro and to initialize this field
  explicitly at the start of the module's initialization function,
  before doing anything else.  This is typically done like this:

\begin{verbatim}
Foo_Type.ob_type = &PyType_Type;
\end{verbatim}

  This should be done before any instances of the type are created.
  \cfunction{PyType_Ready()} checks if \member{ob_type} is \NULL, and
  if so, initializes it: in Python 2.2, it is set to
  \code{\&PyType_Type}; in Python 2.2.1 and later it will be
  initialized to the \member{ob_type} field of the base class.
  \cfunction{PyType_Ready()} will not change this field if it is
  nonzero.

  In Python 2.2, this field is not inherited by subtypes.  In 2.2.1,
  and in 2.3 and beyond, it is inherited by subtypes.
\end{cmemberdesc}

\begin{cmemberdesc}{PyVarObject}{int}{ob_size}
  For statically allocated type objects, this should be initialized
  to zero.  For dynamically allocated type objects, this field has a
  special internal meaning.

  This field is not inherited by subtypes.
\end{cmemberdesc}

\begin{cmemberdesc}{PyTypeObject}{char*}{tp_name}
  Pointer to a NUL-terminated string containing the name of the type.
  For types that are accessible as module globals, the string should
  be the full module name, followed by a dot, followed by the type
  name; for built-in types, it should be just the type name.  If the
  module is a submodule of a package, the full package name is part of
  the full module name.  For example, a type named \class{T} defined
  in module \module{M} in subpackage \module{Q} in package \module{P}
  should have the \member{tp_name} initializer \code{"P.Q.M.T"}.

  For dynamically allocated type objects, this may be just the type
  name, if the module name is explicitly stored in the type dict as
  the value for key \code{'__module__'}.

  If the tp_name field contains a dot, everything before the last dot
  is made accessible as the \member{__module__} attribute, and
  everything after the last dot is made accessible as the
  \member{__name__} attribute.  If no dot is present, the entire
  \member{tp_name} field is made accessible as the \member{__name__}
  attribute, and the \member{__module__} attribute is undefined
  (unless explicitly set in the dictionary, as explained above).

  This field is not inherited by subtypes.
\end{cmemberdesc}

\begin{cmemberdesc}{PyTypeObject}{int}{tp_basicsize}
\cmemberline{PyTypeObject}{int}{tp_itemsize}
  These fields allow calculating the size in byte of instances of
  the type.

  There are two kinds of types: types with fixed-length instances have
  a zero \member{tp_itemsize} field, types with variable-length
  instances have a non-zero \member{tp_itemsize} field.  For a type
  with fixed-length instances, all instances have the same size,
  given in \member{tp_basicsize}.

  For a type with variable-length instances, the instances must have
  an \member{ob_size} field, and the instance size is
  \member{tp_basicsize} plus N times \member{tp_itemsize}, where N is
  the ``length'' of the object.  The value of N is typically stored in
  the instance's \member{ob_size} field.  There are exceptions:  for
  example, long ints use a negative \member{ob_size} to indicate a
  negative number, and N is \code{abs(\member{ob_size})} there.  Also,
  the presence of an \member{ob_size} field in the instance layout
  doesn't mean that the type is variable-length (for example, the list
  type has fixed-length instances, yet those instances have a
  meaningful \member{ob_size} field).

  The basic size includes the fields in the instance declared by the
  macro \csimplemacro{PyObject_HEAD} or
  \csimplemacro{PyObject_VAR_HEAD} (whichever is used to declare the
  instance struct) and this in turn includes the \member{_ob_prev} and
  \member{_ob_next} fields if they are present.  This means that the
  only correct way to get an initializer for the \member{tp_basicsize}
  is to use the \keyword{sizeof} operator on the struct used to
  declare the instance layout.  The basic size does not include the GC
  header size (this is new in Python 2.2; in 2.1 and 2.0, the GC
  header size was included in \member{tp_basicsize}).

  These fields are inherited by subtypes.
\end{cmemberdesc}

\begin{cmemberdesc}{PyTypeObject}{destructor}{tp_dealloc}
  A pointer to the instance destructor function.  This function must
  be defined unless the type guarantees that its instances will never
  be deallocated (as is the case for the singletons \code{None} and
  \code{Ellipsis}).

  The destructor function is called by the \cfunction{Py_DECREF()} and
  \cfunction{Py_XDECREF()} macros when the new reference count is
  zero.  At this point, the instance is still in existance, but there
  are no references to it.  The destructor function should free all
  references which the instance owns, free all memory buffers owned by
  the instance (using the freeing function corresponding to the
  allocation function used to allocate the buffer), and finally (as
  its last action) call the type's \member{tp_free} slot.  If the type
  is not subtypable (doesn't have the \constant{Py_TPFLAGS_BASETYPE}
  flag bit set), it is permissible to call the object deallocator
  directly instead of via \member{tp_free}.  The object deallocator
  should be the one used to allocate the instance; this is normally
  \cfunction{PyObject_Del()} if the instance was allocated using
  \cfunction{PyObject_New()} or \cfunction{PyOject_VarNew()}, or
  \cfunction{PyObject_GC_Del()} if the instance was allocated using
  \cfunction{PyObject_GC_New()} or \cfunction{PyObject_GC_VarNew()}.

  This field is inherited by subtypes.
\end{cmemberdesc}

\begin{cmemberdesc}{PyTypeObject}{printfunc}{tp_print}
  An optional pointer to the instance print function.

  The print function is only called when the instance is printed to a
  \emph{real} file; when it is printed to a pseudo-file (like a
  \class{StringIO} instance), the instance's \member{tp_repr} or
  \member{tp_str} function is called to convert it to a string.  These
  are also called when the type's \member{tp_print} field is \NULL.

  The print function is called with the same signature as
  \cfunction{PyObject_Print()}: \code{tp_print(PyObject *self, FILE
  *file, int flags)}.  The \var{self} argument is the instance to be
  printed.  The \var{file} argument is the stdio file to which it is
  to be printed.  The \var{flags} argument is composed of flag bits.
  The only flag bit currently defined is \constant{Py_PRINT_RAW}.
  When the \constant{Py_PRINT_RAW} flag bit is set, the instance
  should be printed the same way as \member{tp_str} would format it;
  when the \constant{Py_PRINT_RAW} flag bit is clear, the instance
  should be printed the same was as \member{tp_repr} would format it.

  It is possible that the \member{tp_print} field will be deprecated.
  In any case, it is recommended not to define \member{tp_print}, but
  instead to rely on \member{tp_repr} and \member{tp_str} for
  printing.

  This field is inherited by subtypes.
\end{cmemberdesc}

\begin{cmemberdesc}{PyTypeObject}{getattrfunc}{tp_getattr}
  An optional pointer to the get-attribute-string function.

  This field is deprecated.  When it is defined, it should point to a
  function that acts the same as the \member{tp_getattro} function,
  but taking a C string instead of a Python string object to give the
  attribute name.  The signature is the same as for
  \cfunction{PyObject_GetAttrString()}.

  This field is inherited by subtypes together with
  \member{tp_getattro}: a subtype inherits both \member{tp_getattr}
  and \member{tp_getattro} from its base type when the subtype's
  \member{tp_getattr} and \member{tp_getattro} are both \NULL.
\end{cmemberdesc}

\begin{cmemberdesc}{PyTypeObject}{setattrfunc}{tp_setattr}
  An optional pointer to the set-attribute-string function.

  This field is deprecated.  When it is defined, it should point to a
  function that acts the same as the \member{tp_setattro} function,
  but taking a C string instead of a Python string object to give the
  attribute name.  The signature is the same as for
  \cfunction{PyObject_SetAttrString()}.

  This field is inherited by subtypes together with
  \member{tp_setattro}: a subtype inherits both \member{tp_setattr}
  and \member{tp_setattro} from its base type when the subtype's
  \member{tp_setattr} and \member{tp_setattro} are both \NULL.
\end{cmemberdesc}

\begin{cmemberdesc}{PyTypeObject}{cmpfunc}{tp_compare}
  An optional pointer to the three-way comparison function.

  The signature is the same as for \cfunction{PyObject_Compare()}.
  The function should return \code{1} if \var{self} greater than
  \var{other}, \code{0} if \var{self} is equal to \var{other}, and
  \code{-1} if \var{self} less than \var{other}.  It should return
  \code{-1} and set an exception condition when an error occurred
  during the comparison.

  This field is inherited by subtypes together with
  \member{tp_richcompare} and \member{tp_hash}: a subtypes inherits
  all three of \member{tp_compare}, \member{tp_richcompare}, and
  \member{tp_hash} when the subtype's \member{tp_compare},
  \member{tp_richcompare}, and \member{tp_hash} are all \NULL.
\end{cmemberdesc}



\section{Mapping Object Structures \label{mapping-structs}}

\begin{ctypedesc}{PyMappingMethods}
  Structure used to hold pointers to the functions used to implement
  the mapping protocol for an extension type.
\end{ctypedesc}


\section{Number Object Structures \label{number-structs}}

\begin{ctypedesc}{PyNumberMethods}
  Structure used to hold pointers to the functions an extension type
  uses to implement the number protocol.
\end{ctypedesc}


\section{Sequence Object Structures \label{sequence-structs}}

\begin{ctypedesc}{PySequenceMethods}
  Structure used to hold pointers to the functions which an object
  uses to implement the sequence protocol.
\end{ctypedesc}


\section{Buffer Object Structures \label{buffer-structs}}
\sectionauthor{Greg J. Stein}{greg@lyra.org}

The buffer interface exports a model where an object can expose its
internal data as a set of chunks of data, where each chunk is
specified as a pointer/length pair.  These chunks are called
\dfn{segments} and are presumed to be non-contiguous in memory.

If an object does not export the buffer interface, then its
\member{tp_as_buffer} member in the \ctype{PyTypeObject} structure
should be \NULL.  Otherwise, the \member{tp_as_buffer} will point to
a \ctype{PyBufferProcs} structure.

\note{It is very important that your \ctype{PyTypeObject} structure
uses \constant{Py_TPFLAGS_DEFAULT} for the value of the
\member{tp_flags} member rather than \code{0}.  This tells the Python
runtime that your \ctype{PyBufferProcs} structure contains the
\member{bf_getcharbuffer} slot. Older versions of Python did not have
this member, so a new Python interpreter using an old extension needs
to be able to test for its presence before using it.}

\begin{ctypedesc}{PyBufferProcs}
  Structure used to hold the function pointers which define an
  implementation of the buffer protocol.

  The first slot is \member{bf_getreadbuffer}, of type
  \ctype{getreadbufferproc}.  If this slot is \NULL, then the object
  does not support reading from the internal data.  This is
  non-sensical, so implementors should fill this in, but callers
  should test that the slot contains a non-\NULL{} value.

  The next slot is \member{bf_getwritebuffer} having type
  \ctype{getwritebufferproc}.  This slot may be \NULL{} if the object
  does not allow writing into its returned buffers.

  The third slot is \member{bf_getsegcount}, with type
  \ctype{getsegcountproc}.  This slot must not be \NULL{} and is used
  to inform the caller how many segments the object contains.  Simple
  objects such as \ctype{PyString_Type} and \ctype{PyBuffer_Type}
  objects contain a single segment.

  The last slot is \member{bf_getcharbuffer}, of type
  \ctype{getcharbufferproc}.  This slot will only be present if the
  \constant{Py_TPFLAGS_HAVE_GETCHARBUFFER} flag is present in the
  \member{tp_flags} field of the object's \ctype{PyTypeObject}.
  Before using this slot, the caller should test whether it is present
  by using the
  \cfunction{PyType_HasFeature()}\ttindex{PyType_HasFeature()}
  function.  If present, it may be \NULL, indicating that the object's
  contents cannot be used as \emph{8-bit characters}.
  The slot function may also raise an error if the object's contents
  cannot be interpreted as 8-bit characters.  For example, if the
  object is an array which is configured to hold floating point
  values, an exception may be raised if a caller attempts to use
  \member{bf_getcharbuffer} to fetch a sequence of 8-bit characters.
  This notion of exporting the internal buffers as ``text'' is used to
  distinguish between objects that are binary in nature, and those
  which have character-based content.

  \note{The current policy seems to state that these characters
  may be multi-byte characters. This implies that a buffer size of
  \var{N} does not mean there are \var{N} characters present.}
\end{ctypedesc}

\begin{datadesc}{Py_TPFLAGS_HAVE_GETCHARBUFFER}
  Flag bit set in the type structure to indicate that the
  \member{bf_getcharbuffer} slot is known.  This being set does not
  indicate that the object supports the buffer interface or that the
  \member{bf_getcharbuffer} slot is non-\NULL.
\end{datadesc}

\begin{ctypedesc}[getreadbufferproc]{int (*getreadbufferproc)
                            (PyObject *self, int segment, void **ptrptr)}
  Return a pointer to a readable segment of the buffer.  This function
  is allowed to raise an exception, in which case it must return
  \code{-1}.  The \var{segment} which is passed must be zero or
  positive, and strictly less than the number of segments returned by
  the \member{bf_getsegcount} slot function.  On success, it returns
  the length of the buffer memory, and sets \code{*\var{ptrptr}} to a
  pointer to that memory.
\end{ctypedesc}

\begin{ctypedesc}[getwritebufferproc]{int (*getwritebufferproc)
                            (PyObject *self, int segment, void **ptrptr)}
  Return a pointer to a writable memory buffer in
  \code{*\var{ptrptr}}, and the length of that segment as the function
  return value.  The memory buffer must correspond to buffer segment
  \var{segment}.  Must return \code{-1} and set an exception on
  error.  \exception{TypeError} should be raised if the object only
  supports read-only buffers, and \exception{SystemError} should be
  raised when \var{segment} specifies a segment that doesn't exist.
% Why doesn't it raise ValueError for this one?
% GJS: because you shouldn't be calling it with an invalid
%      segment. That indicates a blatant programming error in the C
%      code.
\end{ctypedesc}

\begin{ctypedesc}[getsegcountproc]{int (*getsegcountproc)
                            (PyObject *self, int *lenp)}
  Return the number of memory segments which comprise the buffer.  If
  \var{lenp} is not \NULL, the implementation must report the sum of
  the sizes (in bytes) of all segments in \code{*\var{lenp}}.
  The function cannot fail.
\end{ctypedesc}

\begin{ctypedesc}[getcharbufferproc]{int (*getcharbufferproc)
                            (PyObject *self, int segment, const char **ptrptr)}
\end{ctypedesc}


\section{Supporting the Iterator Protocol
         \label{supporting-iteration}}


\section{Supporting Cyclic Garbarge Collection
         \label{supporting-cycle-detection}}

Python's support for detecting and collecting garbage which involves
circular references requires support from object types which are
``containers'' for other objects which may also be containers.  Types
which do not store references to other objects, or which only store
references to atomic types (such as numbers or strings), do not need
to provide any explicit support for garbage collection.

An example showing the use of these interfaces can be found in
``\ulink{Supporting the Cycle
Collector}{../ext/example-cycle-support.html}'' in
\citetitle[../ext/ext.html]{Extending and Embedding the Python
Interpreter}.

To create a container type, the \member{tp_flags} field of the type
object must include the \constant{Py_TPFLAGS_HAVE_GC} and provide an
implementation of the \member{tp_traverse} handler.  If instances of the
type are mutable, a \member{tp_clear} implementation must also be
provided.

\begin{datadesc}{Py_TPFLAGS_HAVE_GC}
  Objects with a type with this flag set must conform with the rules
  documented here.  For convenience these objects will be referred to
  as container objects.
\end{datadesc}

Constructors for container types must conform to two rules:

\begin{enumerate}
\item  The memory for the object must be allocated using
       \cfunction{PyObject_GC_New()} or \cfunction{PyObject_GC_VarNew()}.

\item  Once all the fields which may contain references to other
       containers are initialized, it must call
       \cfunction{PyObject_GC_Track()}.
\end{enumerate}

\begin{cfuncdesc}{\var{TYPE}*}{PyObject_GC_New}{TYPE, PyTypeObject *type}
  Analogous to \cfunction{PyObject_New()} but for container objects with
  the \constant{Py_TPFLAGS_HAVE_GC} flag set.
\end{cfuncdesc}

\begin{cfuncdesc}{\var{TYPE}*}{PyObject_GC_NewVar}{TYPE, PyTypeObject *type,
                                                   int size}
  Analogous to \cfunction{PyObject_NewVar()} but for container objects
  with the \constant{Py_TPFLAGS_HAVE_GC} flag set.
\end{cfuncdesc}

\begin{cfuncdesc}{PyVarObject *}{PyObject_GC_Resize}{PyVarObject *op, int}
  Resize an object allocated by \cfunction{PyObject_NewVar()}.  Returns
  the resized object or \NULL{} on failure.
\end{cfuncdesc}

\begin{cfuncdesc}{void}{PyObject_GC_Track}{PyObject *op}
  Adds the object \var{op} to the set of container objects tracked by
  the collector.  The collector can run at unexpected times so objects
  must be valid while being tracked.  This should be called once all
  the fields followed by the \member{tp_traverse} handler become valid,
  usually near the end of the constructor.
\end{cfuncdesc}

\begin{cfuncdesc}{void}{_PyObject_GC_TRACK}{PyObject *op}
  A macro version of \cfunction{PyObject_GC_Track()}.  It should not be
  used for extension modules.
\end{cfuncdesc}

Similarly, the deallocator for the object must conform to a similar
pair of rules:

\begin{enumerate}
\item  Before fields which refer to other containers are invalidated,
       \cfunction{PyObject_GC_UnTrack()} must be called.

\item  The object's memory must be deallocated using
       \cfunction{PyObject_GC_Del()}.
\end{enumerate}

\begin{cfuncdesc}{void}{PyObject_GC_Del}{PyObject *op}
  Releases memory allocated to an object using
  \cfunction{PyObject_GC_New()} or \cfunction{PyObject_GC_NewVar()}.
\end{cfuncdesc}

\begin{cfuncdesc}{void}{PyObject_GC_UnTrack}{PyObject *op}
  Remove the object \var{op} from the set of container objects tracked
  by the collector.  Note that \cfunction{PyObject_GC_Track()} can be
  called again on this object to add it back to the set of tracked
  objects.  The deallocator (\member{tp_dealloc} handler) should call
  this for the object before any of the fields used by the
  \member{tp_traverse} handler become invalid.
\end{cfuncdesc}

\begin{cfuncdesc}{void}{_PyObject_GC_UNTRACK}{PyObject *op}
  A macro version of \cfunction{PyObject_GC_UnTrack()}.  It should not be
  used for extension modules.
\end{cfuncdesc}

The \member{tp_traverse} handler accepts a function parameter of this
type:

\begin{ctypedesc}[visitproc]{int (*visitproc)(PyObject *object, void *arg)}
  Type of the visitor function passed to the \member{tp_traverse}
  handler.  The function should be called with an object to traverse
  as \var{object} and the third parameter to the \member{tp_traverse}
  handler as \var{arg}.
\end{ctypedesc}

The \member{tp_traverse} handler must have the following type:

\begin{ctypedesc}[traverseproc]{int (*traverseproc)(PyObject *self,
                                visitproc visit, void *arg)}
  Traversal function for a container object.  Implementations must
  call the \var{visit} function for each object directly contained by
  \var{self}, with the parameters to \var{visit} being the contained
  object and the \var{arg} value passed to the handler.  If
  \var{visit} returns a non-zero value then an error has occurred and
  that value should be returned immediately.
\end{ctypedesc}

The \member{tp_clear} handler must be of the \ctype{inquiry} type, or
\NULL{} if the object is immutable.

\begin{ctypedesc}[inquiry]{int (*inquiry)(PyObject *self)}
  Drop references that may have created reference cycles.  Immutable
  objects do not have to define this method since they can never
  directly create reference cycles.  Note that the object must still
  be valid after calling this method (don't just call
  \cfunction{Py_DECREF()} on a reference).  The collector will call
  this method if it detects that this object is involved in a
  reference cycle.
\end{ctypedesc}