Buffers and Memoryview Objects

Python objects implemented in C can export a group of functions called the “buffer interface.” These functions can be used by an object to expose its data in a raw, byte-oriented format. Clients of the object can use the buffer interface to access the object data directly, without needing to copy it first.

Two examples of objects that support the buffer interface are strings and arrays. The string object exposes the character contents in the buffer interface’s byte-oriented form. An array can only expose its contents via the old-style buffer interface. This limitation does not apply to Python 3, where memoryview objects can be constructed from arrays, too. Array elements may be multi-byte values.

An example user of the buffer interface is the file object’s write() method. Any object that can export a series of bytes through the buffer interface can be written to a file. There are a number of format codes to PyArg_ParseTuple() that operate against an object’s buffer interface, returning data from the target object.

Starting from version 1.6, Python has been providing Python-level buffer objects and a C-level buffer API so that any built-in or used-defined type can expose its characteristics. Both, however, have been deprecated because of various shortcomings, and have been officially removed in Python 3 in favour of a new C-level buffer API and a new Python-level object named memoryview.

The new buffer API has been backported to Python 2.6, and the memoryview object has been backported to Python 2.7. It is strongly advised to use them rather than the old APIs, unless you are blocked from doing so for compatibility reasons.

The new-style Py_buffer struct

type Py_buffer
void *buf

A pointer to the start of the memory for the object.

Py_ssize_t len

The total length of the memory in bytes.

int readonly

An indicator of whether the buffer is read only.

const char *format

A NULL terminated string in struct module style syntax giving the contents of the elements available through the buffer. If this is NULL, "B" (unsigned bytes) is assumed.

int ndim

The number of dimensions the memory represents as a multi-dimensional array. If it is 0, strides and suboffsets must be NULL.

Py_ssize_t *shape

An array of Py_ssize_ts the length of ndim giving the shape of the memory as a multi-dimensional array. Note that ((*shape)[0] * ... * (*shape)[ndims-1])*itemsize should be equal to len.

Py_ssize_t *strides

An array of Py_ssize_ts the length of ndim giving the number of bytes to skip to get to a new element in each dimension.

Py_ssize_t *suboffsets

An array of Py_ssize_ts the length of ndim. If these suboffset numbers are greater than or equal to 0, then the value stored along the indicated dimension is a pointer and the suboffset value dictates how many bytes to add to the pointer after de-referencing. A suboffset value that it 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).

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;
 }
Py_ssize_t itemsize

This is a storage for the itemsize (in bytes) of each element of the shared memory. It is technically un-necessary as it can be obtained using PyBuffer_SizeFromFormat(), however an exporter may know this information without parsing the format string and it is necessary to know the itemsize for proper interpretation of striding. Therefore, storing it is more convenient and faster.

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 should never alter this value.

MemoryView objects

New in version 2.7.

A memoryview object exposes the new C level buffer interface as a Python object which can then be passed around like any other object.

PyObject *PyMemoryView_FromObject(PyObject *obj)

Create a memoryview object from an object that defines the new buffer interface.

PyObject *PyMemoryView_FromBuffer(Py_buffer *view)

Create a memoryview object wrapping the given buffer-info structure view. The memoryview object then owns the buffer, which means you shouldn’t try to release it yourself: it will be released on deallocation of the memoryview object.

PyObject *PyMemoryView_GetContiguous(PyObject *obj, int buffertype, char order)

Create a memoryview object to a contiguous chunk of memory (in either ‘C’ or ‘F’ortran order) from an object that defines the buffer interface. If memory is contiguous, the memoryview object points to the original memory. Otherwise copy is made and the memoryview points to a new bytes object.

int PyMemoryView_Check(PyObject *obj)

Return true if the object obj is a memoryview object. It is not currently allowed to create subclasses of memoryview.

Py_buffer *PyMemoryView_GET_BUFFER(PyObject *obj)

Return a pointer to the buffer-info structure wrapped by the given object. The object must be a memoryview instance; this macro doesn’t check its type, you must do it yourself or you will risk crashes.

Old-style buffer objects

More information on the old buffer interface is provided in the section Buffer Object Structures, under the description for PyBufferProcs.

A “buffer object” is defined in the bufferobject.h header (included by Python.h). These objects look very similar to string objects at the Python programming level: they support slicing, indexing, concatenation, and some other standard string operations. However, their data can come from one of two sources: from a block of memory, or from another object which exports the buffer interface.

Buffer objects are useful as a way to expose the data from another object’s buffer interface 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.

type PyBufferObject

This subtype of PyObject represents a buffer object.

PyTypeObject PyBuffer_Type

The instance of PyTypeObject which represents the Python buffer type; it is the same object as buffer and types.BufferType in the Python layer. .

int Py_END_OF_BUFFER

This constant may be passed as the size parameter to PyBuffer_FromObject() or PyBuffer_FromReadWriteObject(). It indicates that the new PyBufferObject should refer to base object from the specified offset to the end of its exported buffer. Using this enables the caller to avoid querying the base object for its length.

int PyBuffer_Check(PyObject *p)

Return true if the argument has type PyBuffer_Type.

PyObject *PyBuffer_FromObject(PyObject *base, Py_ssize_t offset, Py_ssize_t size)

Return a new read-only buffer object. This raises TypeError if base doesn’t support the read-only buffer protocol or doesn’t provide exactly one buffer segment, or it raises ValueError if offset is less than zero. The buffer will hold a reference to the base object, and the buffer’s contents will refer to the base object’s buffer interface, starting as position offset and extending for size bytes. If size is Py_END_OF_BUFFER, then the new buffer’s contents extend to the length of the base object’s exported buffer data.

Changed in version 2.5: This function used an int type for offset and size. This might require changes in your code for properly supporting 64-bit systems.

PyObject *PyBuffer_FromReadWriteObject(PyObject *base, Py_ssize_t offset, Py_ssize_t size)

Return a new writable buffer object. Parameters and exceptions are similar to those for PyBuffer_FromObject(). If the base object does not export the writeable buffer protocol, then TypeError is raised.

Changed in version 2.5: This function used an int type for offset and size. This might require changes in your code for properly supporting 64-bit systems.

PyObject *PyBuffer_FromMemory(void *ptr, Py_ssize_t size)

Return a new read-only buffer object that reads from a specified location in memory, with a specified size. The caller is responsible for ensuring that the memory buffer, passed in as ptr, is not deallocated while the returned buffer object exists. Raises ValueError if size is less than zero. Note that Py_END_OF_BUFFER may not be passed for the size parameter; ValueError will be raised in that case.

Changed in version 2.5: This function used an int type for size. This might require changes in your code for properly supporting 64-bit systems.

PyObject *PyBuffer_FromReadWriteMemory(void *ptr, Py_ssize_t size)

Similar to PyBuffer_FromMemory(), but the returned buffer is writable.

Changed in version 2.5: This function used an int type for size. This might require changes in your code for properly supporting 64-bit systems.

PyObject *PyBuffer_New(Py_ssize_t size)

Return a new writable buffer object that maintains its own memory buffer of size bytes. ValueError is returned if size is not zero or positive. Note that the memory buffer (as returned by PyObject_AsWriteBuffer()) is not specifically aligned.

Changed in version 2.5: This function used an int type for size. This might require changes in your code for properly supporting 64-bit systems.