Doc/c-api/buffer.rst

.. highlightlang:: c

.. index::
   single: buffer protocol
   single: buffer interface; (see buffer protocol)
   single: buffer object; (see buffer protocol)

.. _bufferobjects:

Buffer Protocol
---------------

.. sectionauthor:: Greg Stein <gstein@lyra.org>
.. sectionauthor:: Benjamin Peterson
.. sectionauthor:: Stefan Krah


Certain objects available in Python wrap access to an underlying memory
array or *buffer*.  Such objects include the built-in :class:`bytes` and
:class:`bytearray`, and some extension types like :class:`array.array`.
Third-party libraries may define their own types for special purposes, such
as image processing or numeric analysis.

While each of these types have their own semantics, they share the common
characteristic of being backed by a possibly large memory buffer.  It is
then desirable, in some situations, to access that buffer directly and
without intermediate copying.

Python provides such a facility at the C level in the form of the :ref:`buffer
protocol <bufferobjects>`.  This protocol has two sides:

.. index:: single: PyBufferProcs

- on the producer side, a type can export a "buffer interface" which allows
  objects of that type to expose information about their underlying buffer.
  This interface is described in the section :ref:`buffer-structs`;

- on the consumer side, several means are available to obtain a pointer to
  the raw underlying data of an object (for example a method parameter).

Simple objects such as :class:`bytes` and :class:`bytearray` expose their
underlying buffer in byte-oriented form.  Other forms are possible; for example,
the elements exposed by a :class:`array.array` can be multi-byte values.

An example consumer of the buffer interface is the :meth:`~io.BufferedIOBase.write`
method of file objects: any object that can export a series of bytes through
the buffer interface can be written to a file.  While :meth:`write` only
needs read-only access to the internal contents of the object passed to it,
other methods such as :meth:`~io.BufferedIOBase.readinto` need write access
to the contents of their argument.  The buffer interface allows objects to
selectively allow or reject exporting of read-write and read-only buffers.

There are two ways for a consumer of the buffer interface to acquire a buffer
over a target object:

* call :c:func:`PyObject_GetBuffer` with the right parameters;

* call :c:func:`PyArg_ParseTuple` (or one of its siblings) with one of the
  ``y*``, ``w*`` or ``s*`` :ref:`format codes <arg-parsing>`.

In both cases, :c:func:`PyBuffer_Release` must be called when the buffer
isn't needed anymore.  Failure to do so could lead to various issues such as
resource leaks.


.. _buffer-structure:

Buffer structure
================

Buffer structures (or simply "buffers") are useful as a way to expose the
binary data from another object to the Python programmer.  They can also be
used as a zero-copy slicing mechanism.  Using their ability to reference a
block of memory, it is possible to expose any data to the Python programmer
quite easily.  The memory could be a large, constant array in a C extension,
it could be a raw block of memory for manipulation before passing to an
operating system library, or it could be used to pass around structured data
in its native, in-memory format.

Contrary to most data types exposed by the Python interpreter, buffers
are not :c:type:`PyObject` pointers but rather simple C structures.  This
allows them to be created and copied very simply.  When a generic wrapper
around a buffer is needed, a :ref:`memoryview <memoryview-objects>` object
can be created.

For short instructions how to write an exporting object, see
:ref:`Buffer Object Structures <buffer-structs>`. For obtaining
a buffer, see :c:func:`PyObject_GetBuffer`.

.. c:type:: Py_buffer

   .. c:member:: void \*buf

      A pointer to the start of the logical structure described by the buffer
      fields. This can be any location within the underlying physical memory
      block of the exporter. For example, with negative :c:member:`~Py_buffer.strides`
      the value may point to the end of the memory block.

      For contiguous arrays, the value points to the beginning of the memory
      block.

   .. c:member:: void \*obj

      A new reference to the exporting object. The reference is owned by
      the consumer and automatically decremented and set to *NULL* by
      :c:func:`PyBuffer_Release`. The field is the equivalent of the return
      value of any standard C-API function.

      As a special case, for *temporary* buffers that are wrapped by
      :c:func:`PyMemoryView_FromBuffer` or :c:func:`PyBuffer_FillInfo`
      this field is *NULL*. In general, exporting objects MUST NOT
      use this scheme.

   .. c:member:: Py_ssize_t len

      ``product(shape) * itemsize``. For contiguous arrays, this is the length
      of the underlying memory block. For non-contiguous arrays, it is the length
      that the logical structure would have if it were copied to a contiguous
      representation.

      Accessing ``((char *)buf)[0] up to ((char *)buf)[len-1]`` is only valid
      if the buffer has been obtained by a request that guarantees contiguity. In
      most cases such a request will be :c:macro:`PyBUF_SIMPLE` or :c:macro:`PyBUF_WRITABLE`.

   .. c:member:: int readonly

      An indicator of whether the buffer is read-only. This field is controlled
      by the :c:macro:`PyBUF_WRITABLE` flag.

   .. c:member:: Py_ssize_t itemsize

      Item size in bytes of a single element. Same as the value of :func:`struct.calcsize`
      called on non-NULL :c:member:`~Py_buffer.format` values.

      Important exception: If a consumer requests a buffer without the
      :c:macro:`PyBUF_FORMAT` flag, :c:member:`~Py_buffer.format` will
      be set to  *NULL*,  but :c:member:`~Py_buffer.itemsize` still has
      the value for the original format.

      If :c:member:`~Py_buffer.shape` is present, the equality
      ``product(shape) * itemsize == len`` still holds and the consumer
      can use :c:member:`~Py_buffer.itemsize` to navigate the buffer.

      If :c:member:`~Py_buffer.shape` is *NULL* as a result of a :c:macro:`PyBUF_SIMPLE`
      or a :c:macro:`PyBUF_WRITABLE` request, the consumer must disregard
      :c:member:`~Py_buffer.itemsize` and assume ``itemsize == 1``.

   .. c:member:: const char \*format

      A *NUL* terminated string in :mod:`struct` module style syntax describing
      the contents of a single item. If this is *NULL*, ``"B"`` (unsigned bytes)
      is assumed.

      This field is controlled by the :c:macro:`PyBUF_FORMAT` flag.

   .. c:member:: int ndim

      The number of dimensions the memory represents as an n-dimensional array.
      If it is 0, :c:member:`~Py_buffer.buf` points to a single item representing
      a scalar. In this case, :c:member:`~Py_buffer.shape`, :c:member:`~Py_buffer.strides`
      and :c:member:`~Py_buffer.suboffsets` MUST be *NULL*.

      The macro :c:macro:`PyBUF_MAX_NDIM` limits the maximum number of dimensions
      to 64. Exporters MUST respect this limit, consumers of multi-dimensional
      buffers SHOULD be able to handle up to :c:macro:`PyBUF_MAX_NDIM` dimensions.

   .. c:member:: Py_ssize_t \*shape

      An array of :c:type:`Py_ssize_t` of length :c:member:`~Py_buffer.ndim`
      indicating the shape of the memory as an n-dimensional array. Note that
      ``shape[0] * ... * shape[ndim-1] * itemsize`` MUST be equal to
      :c:member:`~Py_buffer.len`.

      Shape values are restricted to ``shape[n] >= 0``. The case
      ``shape[n] == 0`` requires special attention. See `complex arrays`_
      for further information.

      The shape array is read-only for the consumer.

   .. c:member:: Py_ssize_t \*strides

      An array of :c:type:`Py_ssize_t` of length :c:member:`~Py_buffer.ndim`
      giving the number of bytes to skip to get to a new element in each
      dimension.

      Stride values can be any integer. For regular arrays, strides are
      usually positive, but a consumer MUST be able to handle the case
      ``strides[n] <= 0``. See `complex arrays`_ for further information.

      The strides array is read-only for the consumer.

   .. c:member:: Py_ssize_t \*suboffsets

      An array of :c:type:`Py_ssize_t` of length :c:member:`~Py_buffer.ndim`.
      If ``suboffsets[n] >= 0``, the values stored along the nth dimension are
      pointers and the suboffset value dictates how many bytes to add to each
      pointer after de-referencing. A suboffset value that is negative
      indicates that no de-referencing should occur (striding in a contiguous
      memory block).

      If all suboffsets are negative (i.e. no de-referencing is needed, then
      this field must be NULL (the default value).

      This type of array representation is used by the Python Imaging Library
      (PIL). See `complex arrays`_ for further information how to access elements
      of such an array.

      The suboffsets array is read-only for the consumer.

   .. c:member:: void \*internal

      This is for use internally by the exporting object. For example, this
      might be re-cast as an integer by the exporter and used to store flags
      about whether or not the shape, strides, and suboffsets arrays must be
      freed when the buffer is released. The consumer MUST NOT alter this
      value.

.. _buffer-request-types:

Buffer request types
====================

Buffers are usually obtained by sending a buffer request to an exporting
object via :c:func:`PyObject_GetBuffer`. Since the complexity of the logical
structure of the memory can vary drastically, the consumer uses the *flags*
argument to specify the exact buffer type it can handle.

All :c:data:`Py_buffer` fields are unambiguously defined by the request
type.

request-independent fields
~~~~~~~~~~~~~~~~~~~~~~~~~~
The following fields are not influenced by *flags* and must always be filled in
with the correct values: :c:member:`~Py_buffer.obj`, :c:member:`~Py_buffer.buf`,
:c:member:`~Py_buffer.len`, :c:member:`~Py_buffer.itemsize`, :c:member:`~Py_buffer.ndim`.


readonly, format
~~~~~~~~~~~~~~~~

   .. c:macro:: PyBUF_WRITABLE

      Controls the :c:member:`~Py_buffer.readonly` field. If set, the exporter
      MUST provide a writable buffer or else report failure. Otherwise, the
      exporter MAY provide either a read-only or writable buffer, but the choice
      MUST be consistent for all consumers.

   .. c:macro:: PyBUF_FORMAT

      Controls the :c:member:`~Py_buffer.format` field. If set, this field MUST
      be filled in correctly. Otherwise, this field MUST be *NULL*.


:c:macro:`PyBUF_WRITABLE` can be \|'d to any of the flags in the next section.
Since :c:macro:`PyBUF_SIMPLE` is defined as 0, :c:macro:`PyBUF_WRITABLE`
can be used as a stand-alone flag to request a simple writable buffer.

:c:macro:`PyBUF_FORMAT` can be \|'d to any of the flags except :c:macro:`PyBUF_SIMPLE`.
The latter already implies format ``B`` (unsigned bytes).


shape, strides, suboffsets
~~~~~~~~~~~~~~~~~~~~~~~~~~

The flags that control the logical structure of the memory are listed
in decreasing order of complexity. Note that each flag contains all bits
of the flags below it.

.. tabularcolumns:: |p{0.35\linewidth}|l|l|l|

+-----------------------------+-------+---------+------------+
|  Request                    | shape | strides | suboffsets |
+=============================+=======+=========+============+
| .. c:macro:: PyBUF_INDIRECT |  yes  |   yes   | if needed  |
+-----------------------------+-------+---------+------------+
| .. c:macro:: PyBUF_STRIDES  |  yes  |   yes   |    NULL    |
+-----------------------------+-------+---------+------------+
| .. c:macro:: PyBUF_ND       |  yes  |   NULL  |    NULL    |
+-----------------------------+-------+---------+------------+
| .. c:macro:: PyBUF_SIMPLE   |  NULL |   NULL  |    NULL    |
+-----------------------------+-------+---------+------------+


contiguity requests
~~~~~~~~~~~~~~~~~~~

C or Fortran contiguity can be explicitly requested, with and without stride
information. Without stride information, the buffer must be C-contiguous.

.. tabularcolumns:: |p{0.35\linewidth}|l|l|l|l|

+-----------------------------------+-------+---------+------------+--------+
|  Request                          | shape | strides | suboffsets | contig |
+===================================+=======+=========+============+========+
| .. c:macro:: PyBUF_C_CONTIGUOUS   |  yes  |   yes   |    NULL    |   C    |
+-----------------------------------+-------+---------+------------+--------+
| .. c:macro:: PyBUF_F_CONTIGUOUS   |  yes  |   yes   |    NULL    |   F    |
+-----------------------------------+-------+---------+------------+--------+
| .. c:macro:: PyBUF_ANY_CONTIGUOUS |  yes  |   yes   |    NULL    | C or F |
+-----------------------------------+-------+---------+------------+--------+
| .. c:macro:: PyBUF_ND             |  yes  |   NULL  |    NULL    |   C    |
+-----------------------------------+-------+---------+------------+--------+


compound requests
~~~~~~~~~~~~~~~~~

All possible requests are fully defined by some combination of the flags in
the previous section. For convenience, the buffer protocol provides frequently
used combinations as single flags.

In the following table *U* stands for undefined contiguity. The consumer would
have to call :c:func:`PyBuffer_IsContiguous` to determine contiguity.

.. tabularcolumns:: |p{0.35\linewidth}|l|l|l|l|l|l|

+-------------------------------+-------+---------+------------+--------+----------+--------+
|  Request                      | shape | strides | suboffsets | contig | readonly | format |
+===============================+=======+=========+============+========+==========+========+
| .. c:macro:: PyBUF_FULL       |  yes  |   yes   | if needed  |   U    |     0    |  yes   |
+-------------------------------+-------+---------+------------+--------+----------+--------+
| .. c:macro:: PyBUF_FULL_RO    |  yes  |   yes   | if needed  |   U    |  1 or 0  |  yes   |
+-------------------------------+-------+---------+------------+--------+----------+--------+
| .. c:macro:: PyBUF_RECORDS    |  yes  |   yes   |    NULL    |   U    |     0    |  yes   |
+-------------------------------+-------+---------+------------+--------+----------+--------+
| .. c:macro:: PyBUF_RECORDS_RO |  yes  |   yes   |    NULL    |   U    |  1 or 0  |  yes   |
+-------------------------------+-------+---------+------------+--------+----------+--------+
| .. c:macro:: PyBUF_STRIDED    |  yes  |   yes   |    NULL    |   U    |     0    |  NULL  |
+-------------------------------+-------+---------+------------+--------+----------+--------+
| .. c:macro:: PyBUF_STRIDED_RO |  yes  |   yes   |    NULL    |   U    |  1 or 0  |  NULL  |
+-------------------------------+-------+---------+------------+--------+----------+--------+
| .. c:macro:: PyBUF_CONTIG     |  yes  |   NULL  |    NULL    |   C    |     0    |  NULL  |
+-------------------------------+-------+---------+------------+--------+----------+--------+
| .. c:macro:: PyBUF_CONTIG_RO  |  yes  |   NULL  |    NULL    |   C    |  1 or 0  |  NULL  |
+-------------------------------+-------+---------+------------+--------+----------+--------+


Complex arrays
==============

NumPy-style: shape and strides
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

The logical structure of NumPy-style arrays is defined by :c:member:`~Py_buffer.itemsize`,
:c:member:`~Py_buffer.ndim`, :c:member:`~Py_buffer.shape` and :c:member:`~Py_buffer.strides`.

If ``ndim == 0``, the memory location pointed to by :c:member:`~Py_buffer.buf` is
interpreted as a scalar of size :c:member:`~Py_buffer.itemsize`. In that case,
both :c:member:`~Py_buffer.shape` and :c:member:`~Py_buffer.strides` are *NULL*.

If :c:member:`~Py_buffer.strides` is *NULL*, the array is interpreted as
a standard n-dimensional C-array. Otherwise, the consumer must access an
n-dimensional array as follows:

   ``ptr = (char *)buf + indices[0] * strides[0] + ... + indices[n-1] * strides[n-1]``
   ``item = *((typeof(item) *)ptr);``


As noted above, :c:member:`~Py_buffer.buf` can point to any location within
the actual memory block. An exporter can check the validity of a buffer with
this function:

.. code-block:: python

   def verify_structure(memlen, itemsize, ndim, shape, strides, offset):
       """Verify that the parameters represent a valid array within
          the bounds of the allocated memory:
              char *mem: start of the physical memory block
              memlen: length of the physical memory block
              offset: (char *)buf - mem
       """
       if offset % itemsize:
           return False
       if offset < 0 or offset+itemsize > memlen:
           return False
       if any(v % itemsize for v in strides):
           return False

       if ndim <= 0:
           return ndim == 0 and not shape and not strides
       if 0 in shape:
           return True

       imin = sum(strides[j]*(shape[j]-1) for j in range(ndim)
                  if strides[j] <= 0)
       imax = sum(strides[j]*(shape[j]-1) for j in range(ndim)
                  if strides[j] > 0)

       return 0 <= offset+imin and offset+imax+itemsize <= memlen


PIL-style: shape, strides and suboffsets
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

In addition to the regular items, PIL-style arrays can contain pointers
that must be followed in order to get to the next element in a dimension.
For example, the regular three-dimensional C-array ``char v[2][2][3]`` can
also be viewed as an array of 2 pointers to 2 two-dimensional arrays:
``char (*v[2])[2][3]``. In suboffsets representation, those two pointers
can be embedded at the start of :c:member:`~Py_buffer.buf`, pointing
to two ``char x[2][3]`` arrays that can be located anywhere in memory.


Here is a function that returns a pointer to the element in an N-D array
pointed to by an N-dimensional index when there are both non-NULL strides
and suboffsets::

   void *get_item_pointer(int ndim, void *buf, Py_ssize_t *strides,
                          Py_ssize_t *suboffsets, Py_ssize_t *indices) {
       char *pointer = (char*)buf;
       int i;
       for (i = 0; i < ndim; i++) {
           pointer += strides[i] * indices[i];
           if (suboffsets[i] >=0 ) {
               pointer = *((char**)pointer) + suboffsets[i];
           }
       }
       return (void*)pointer;
   }


Buffer-related functions
========================

.. c:function:: int PyObject_CheckBuffer(PyObject *obj)

   Return 1 if *obj* supports the buffer interface otherwise 0.  When 1 is
   returned, it doesn't guarantee that :c:func:`PyObject_GetBuffer` will
   succeed.


.. c:function:: int PyObject_GetBuffer(PyObject *exporter, Py_buffer *view, int flags)

   Send a request to *exporter* to fill in *view* as specified by  *flags*.
   If the exporter cannot provide a buffer of the exact type, it MUST raise
   :c:data:`PyExc_BufferError`, set :c:member:`view->obj` to *NULL* and
   return -1.

   On success, fill in *view*, set :c:member:`view->obj` to a new reference
   to *exporter* and return 0. In the case of chained buffer providers
   that redirect requests to a single object, :c:member:`view->obj` MAY
   refer to this object instead of *exporter* (See :ref:`Buffer Object Structures <buffer-structs>`).

   Successful calls to :c:func:`PyObject_GetBuffer` must be paired with calls
   to :c:func:`PyBuffer_Release`, similar to :c:func:`malloc` and :c:func:`free`.
   Thus, after the consumer is done with the buffer, :c:func:`PyBuffer_Release`
   must be called exactly once.


.. c:function:: void PyBuffer_Release(Py_buffer *view)

   Release the buffer *view* and decrement the reference count for
   :c:member:`view->obj`. This function MUST be called when the buffer
   is no longer being used, otherwise reference leaks may occur.

   It is an error to call this function on a buffer that was not obtained via
   :c:func:`PyObject_GetBuffer`.


.. c:function:: Py_ssize_t PyBuffer_SizeFromFormat(const char *)

   Return the implied :c:data:`~Py_buffer.itemsize` from :c:data:`~Py_buffer.format`.
   This function is not yet implemented.


.. c:function:: int PyBuffer_IsContiguous(Py_buffer *view, char order)

   Return 1 if the memory defined by the *view* is C-style (*order* is
   ``'C'``) or Fortran-style (*order* is ``'F'``) contiguous or either one
   (*order* is ``'A'``).  Return 0 otherwise.


.. c:function:: void PyBuffer_FillContiguousStrides(int ndim, Py_ssize_t *shape, Py_ssize_t *strides, Py_ssize_t itemsize, char order)

   Fill the *strides* array with byte-strides of a contiguous (C-style if
   *order* is ``'C'`` or Fortran-style if *order* is ``'F'``) array of the
   given shape with the given number of bytes per element.


.. c:function:: int PyBuffer_FillInfo(Py_buffer *view, PyObject *exporter, void *buf, Py_ssize_t len, int readonly, int flags)

   Handle buffer requests for an exporter that wants to expose *buf* of size *len*
   with writability set according to *readonly*. *buf* is interpreted as a sequence
   of unsigned bytes.

   The *flags* argument indicates the request type. This function always fills in
   *view* as specified by flags, unless *buf* has been designated as read-only
   and :c:macro:`PyBUF_WRITABLE` is set in *flags*.

   On success, set :c:member:`view->obj` to a new reference to *exporter* and
   return 0. Otherwise, raise :c:data:`PyExc_BufferError`, set
   :c:member:`view->obj` to *NULL* and return -1;

   If this function is used as part of a :ref:`getbufferproc <buffer-structs>`,
   *exporter* MUST be set to the exporting object and *flags* must be passed
   unmodified. Otherwise, *exporter* MUST be NULL.