Buffer Protocol¶
Certain objects available in Python wrap access to an underlying memoryarray orbuffer. Such objects include the built-inbytes
andbytearray
, and some extension types likearray.array
.Third-party libraries may define their own types for special purposes, suchas image processing or numeric analysis.
While each of these types have their own semantics, they share the commoncharacteristic of being backed by a possibly large memory buffer. It isthen desirable, in some situations, to access that buffer directly andwithout intermediate copying.
Python provides such a facility at the C level in the form of thebufferprotocol. This protocol has two sides:
on the producer side, a type can export a “buffer interface” which allowsobjects of that type to expose information about their underlying buffer.This interface is described in the sectionBuffer Object Structures;
on the consumer side, several means are available to obtain a pointer tothe raw underlying data of an object (for example a method parameter).
Simple objects such asbytes
andbytearray
expose theirunderlying buffer in byte-oriented form. Other forms are possible; for example,the elements exposed by anarray.array
can be multi-byte values.
An example consumer of the buffer interface is thewrite()
method of file objects: any object that can export a series of bytes throughthe buffer interface can be written to a file. Whilewrite()
onlyneeds read-only access to the internal contents of the object passed to it,other methods such asreadinto()
need write accessto the contents of their argument. The buffer interface allows objects toselectively 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 bufferover a target object:
call
PyObject_GetBuffer()
with the right parameters;call
PyArg_ParseTuple()
(or one of its siblings) with one of they*
,w*
ors*
format codes.
In both cases,PyBuffer_Release()
must be called when the bufferisn’t needed anymore. Failure to do so could lead to various issues such asresource leaks.
Buffer structure¶
Buffer structures (or simply “buffers”) are useful as a way to expose thebinary data from another object to the Python programmer. They can also beused as a zero-copy slicing mechanism. Using their ability to reference ablock of memory, it is possible to expose any data to the Python programmerquite 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 anoperating system library, or it could be used to pass around structured datain its native, in-memory format.
Contrary to most data types exposed by the Python interpreter, buffersare notPyObject
pointers but rather simple C structures. Thisallows them to be created and copied very simply. When a generic wrapperaround a buffer is needed, amemoryview objectcan be created.
For short instructions how to write an exporting object, seeBuffer Object Structures. For obtaininga buffer, seePyObject_GetBuffer()
.
Py_buffer
¶- void *
buf
¶ A pointer to the start of the logical structure described by the bufferfields. This can be any location within the underlying physical memoryblock of the exporter. For example, with negative
strides
the value may point to the end of the memory block.Forcontiguous arrays, the value points to the beginning ofthe memory block.
- void *
obj
¶ A new reference to the exporting object. The reference is owned bythe consumer and automatically decremented and set to
NULL
byPyBuffer_Release()
. The field is the equivalent of the returnvalue of any standard C-API function.As a special case, fortemporary buffers that are wrapped by
PyMemoryView_FromBuffer()
orPyBuffer_FillInfo()
this field isNULL
. In general, exporting objects MUST NOTuse this scheme.
- Py_ssize_t
len
¶ product(shape)*itemsize
. For contiguous arrays, this is the lengthof the underlying memory block. For non-contiguous arrays, it is the lengththat the logical structure would have if it were copied to a contiguousrepresentation.Accessing
((char*)buf)[0]upto((char*)buf)[len-1]
is only validif the buffer has been obtained by a request that guarantees contiguity. Inmost cases such a request will bePyBUF_SIMPLE
orPyBUF_WRITABLE
.
- int
readonly
¶ An indicator of whether the buffer is read-only. This field is controlledby the
PyBUF_WRITABLE
flag.
- Py_ssize_t
itemsize
¶ Item size in bytes of a single element. Same as the value of
struct.calcsize()
called on non-NULL
format
values.Important exception: If a consumer requests a buffer without the
PyBUF_FORMAT
flag,format
willbe set toNULL
, butitemsize
still hasthe value for the original format.If
shape
is present, the equalityproduct(shape)*itemsize==len
still holds and the consumercan useitemsize
to navigate the buffer.If
shape
isNULL
as a result of aPyBUF_SIMPLE
or aPyBUF_WRITABLE
request, the consumer must disregarditemsize
and assumeitemsize==1
.
- const char *
format
¶ ANUL terminated string in
struct
module style syntax describingthe contents of a single item. If this isNULL
,"B"
(unsigned bytes)is assumed.This field is controlled by the
PyBUF_FORMAT
flag.
- int
ndim
¶ The number of dimensions the memory represents as an n-dimensional array.If it is
0
,buf
points to a single item representinga scalar. In this case,shape
,strides
andsuboffsets
MUST beNULL
.The macro
PyBUF_MAX_NDIM
limits the maximum number of dimensionsto 64. Exporters MUST respect this limit, consumers of multi-dimensionalbuffers SHOULD be able to handle up toPyBUF_MAX_NDIM
dimensions.
- Py_ssize_t *
shape
¶ An array of
Py_ssize_t
of lengthndim
indicating the shape of the memory as an n-dimensional array. Note thatshape[0]*...*shape[ndim-1]*itemsize
MUST be equal tolen
.Shape values are restricted to
shape[n]>=0
. The caseshape[n]==0
requires special attention. Seecomplex arraysfor further information.The shape array is read-only for the consumer.
- Py_ssize_t *
strides
¶ An array of
Py_ssize_t
of lengthndim
giving the number of bytes to skip to get to a new element in eachdimension.Stride values can be any integer. For regular arrays, strides areusually positive, but a consumer MUST be able to handle the case
strides[n]<=0
. Seecomplex arrays for further information.The strides array is read-only for the consumer.
- Py_ssize_t *
suboffsets
¶ An array of
Py_ssize_t
of lengthndim
.Ifsuboffsets[n]>=0
, the values stored along the nth dimension arepointers and the suboffset value dictates how many bytes to add to eachpointer after de-referencing. A suboffset value that is negativeindicates that no de-referencing should occur (striding in a contiguousmemory block).If all suboffsets are negative (i.e. no de-referencing is needed), thenthis field must be
NULL
(the default value).This type of array representation is used by the Python Imaging Library(PIL). Seecomplex arrays for further information how to access elementsof such an array.
The suboffsets array is read-only for the consumer.
- void *
internal
¶ This is for use internally by the exporting object. For example, thismight be re-cast as an integer by the exporter and used to store flagsabout whether or not the shape, strides, and suboffsets arrays must befreed when the buffer is released. The consumer MUST NOT alter thisvalue.
- void *
Buffer request types¶
Buffers are usually obtained by sending a buffer request to an exportingobject viaPyObject_GetBuffer()
. Since the complexity of the logicalstructure of the memory can vary drastically, the consumer uses theflagsargument to specify the exact buffer type it can handle.
AllPy_buffer
fields are unambiguously defined by the requesttype.
request-independent fields¶
The following fields are not influenced byflags and must always be filled inwith the correct values:obj
,buf
,len
,itemsize
,ndim
.
readonly, format¶
PyBUF_WRITABLE
can be |’d to any of the flags in the next section.SincePyBUF_SIMPLE
is defined as 0,PyBUF_WRITABLE
can be used as a stand-alone flag to request a simple writable buffer.
PyBUF_FORMAT
can be |’d to any of the flags exceptPyBUF_SIMPLE
.The latter already implies formatB
(unsigned bytes).
shape, strides, suboffsets¶
The flags that control the logical structure of the memory are listedin decreasing order of complexity. Note that each flag contains all bitsof the flags below it.
Request | shape | strides | suboffsets |
---|---|---|---|
| yes | yes | if needed |
| yes | yes | NULL |
| yes | NULL | NULL |
| NULL | NULL | NULL |
contiguity requests¶
C or Fortrancontiguity can be explicitly requested,with and without stride information. Without stride information, the buffermust be C-contiguous.
Request | shape | strides | suboffsets | contig |
---|---|---|---|---|
| yes | yes | NULL | C |
| yes | yes | NULL | F |
| yes | yes | NULL | C or F |
yes | NULL | NULL | C |
compound requests¶
All possible requests are fully defined by some combination of the flags inthe previous section. For convenience, the buffer protocol provides frequentlyused combinations as single flags.
In the following tableU stands for undefined contiguity. The consumer wouldhave to callPyBuffer_IsContiguous()
to determine contiguity.
Request | shape | strides | suboffsets | contig | readonly | format |
---|---|---|---|---|---|---|
| yes | yes | if needed | U | 0 | yes |
| yes | yes | if needed | U | 1 or 0 | yes |
| yes | yes | NULL | U | 0 | yes |
| yes | yes | NULL | U | 1 or 0 | yes |
| yes | yes | NULL | U | 0 | NULL |
| yes | yes | NULL | U | 1 or 0 | NULL |
| yes | NULL | NULL | C | 0 | NULL |
| yes | NULL | NULL | C | 1 or 0 | NULL |
Complex arrays¶
NumPy-style: shape and strides¶
The logical structure of NumPy-style arrays is defined byitemsize
,ndim
,shape
andstrides
.
Ifndim==0
, the memory location pointed to bybuf
isinterpreted as a scalar of sizeitemsize
. In that case,bothshape
andstrides
areNULL
.
Ifstrides
isNULL
, the array is interpreted asa standard n-dimensional C-array. Otherwise, the consumer must access ann-dimensional array as follows:
ptr=(char*)buf+indices[0]*strides[0]+...+indices[n-1]*strides[n-1];item=*((typeof(item)*)ptr);
As noted above,buf
can point to any location withinthe actual memory block. An exporter can check the validity of a buffer withthis function:
defverify_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 """ifoffset%itemsize:returnFalseifoffset<0oroffset+itemsize>memlen:returnFalseifany(v%itemsizeforvinstrides):returnFalseifndim<=0:returnndim==0andnotshapeandnotstridesif0inshape:returnTrueimin=sum(strides[j]*(shape[j]-1)forjinrange(ndim)ifstrides[j]<=0)imax=sum(strides[j]*(shape[j]-1)forjinrange(ndim)ifstrides[j]>0)return0<=offset+iminandoffset+imax+itemsize<=memlen
PIL-style: shape, strides and suboffsets¶
In addition to the regular items, PIL-style arrays can contain pointersthat must be followed in order to get to the next element in a dimension.For example, the regular three-dimensional C-arraycharv[2][2][3]
canalso be viewed as an array of 2 pointers to 2 two-dimensional arrays:char(*v[2])[2][3]
. In suboffsets representation, those two pointerscan be embedded at the start ofbuf
, pointingto twocharx[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 arraypointed to by an N-dimensional index when there are both non-NULL
stridesand suboffsets:
void*get_item_pointer(intndim,void*buf,Py_ssize_t*strides,Py_ssize_t*suboffsets,Py_ssize_t*indices){char*pointer=(char*)buf;inti;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¶
- int
PyObject_CheckBuffer
(PyObject *obj)¶ Return
1
ifobj supports the buffer interface otherwise0
. When1
isreturned, it doesn’t guarantee thatPyObject_GetBuffer()
willsucceed. This function always succeeds.
- int
PyObject_GetBuffer
(PyObject *exporter,Py_buffer *view, int flags)¶ Send a request toexporter to fill inview as specified byflags.If the exporter cannot provide a buffer of the exact type, it MUST raise
PyExc_BufferError
, setview->obj
toNULL
andreturn-1
.On success, fill inview, set
view->obj
to a new referencetoexporter and return 0. In the case of chained buffer providersthat redirect requests to a single object,view->obj
MAYrefer to this object instead ofexporter (SeeBuffer Object Structures).Successful calls to
PyObject_GetBuffer()
must be paired with callstoPyBuffer_Release()
, similar tomalloc()
andfree()
.Thus, after the consumer is done with the buffer,PyBuffer_Release()
must be called exactly once.
- void
PyBuffer_Release
(Py_buffer *view)¶ Release the bufferview and decrement the reference count for
view->obj
. This function MUST be called when the bufferis 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
PyObject_GetBuffer()
.
- Py_ssize_t
PyBuffer_SizeFromFormat
(const char *format)¶ Return the implied
itemsize
fromformat
.On error, raise an exception and return -1.New in version 3.9.
- int
PyBuffer_IsContiguous
(Py_buffer *view, char order)¶ Return
1
if the memory defined by theview is C-style (order is'C'
) or Fortran-style (order is'F'
)contiguous or either one(order is'A'
). Return0
otherwise. This function always succeeds.
- void*
PyBuffer_GetPointer
(Py_buffer *view,Py_ssize_t *indices)¶ Get the memory area pointed to by theindices inside the givenview.indices must point to an array of
view->ndim
indices.
- int
PyBuffer_FromContiguous
(Py_buffer *view, void *buf,Py_ssize_t len, char fort)¶ Copy contiguouslen bytes frombuf toview.fort can be
'C'
or'F'
(for C-style or Fortran-style ordering).0
is returned on success,-1
on error.
- int
PyBuffer_ToContiguous
(void *buf,Py_buffer *src,Py_ssize_t len, char order)¶ Copylen bytes fromsrc to its contiguous representation inbuf.order can be
'C'
or'F'
or'A'
(for C-style or Fortran-styleordering or either one).0
is returned on success,-1
on error.This function fails iflen !=src->len.
- void
PyBuffer_FillContiguousStrides
(int ndims,Py_ssize_t *shape,Py_ssize_t *strides, int itemsize, char order)¶ Fill thestrides array with byte-strides of acontiguous (C-style iforder is
'C'
or Fortran-style iforder is'F'
) array of thegiven shape with the given number of bytes per element.
- 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 exposebuf of sizelenwith writability set according toreadonly.buf is interpreted as a sequenceof unsigned bytes.
Theflags argument indicates the request type. This function always fills inview as specified by flags, unlessbuf has been designated as read-onlyand
PyBUF_WRITABLE
is set inflags.On success, set
view->obj
to a new reference toexporter andreturn 0. Otherwise, raisePyExc_BufferError
, setview->obj
toNULL
and return-1
;If this function is used as part of agetbufferproc,exporter MUST be set to the exporting object andflags must be passedunmodified. Otherwise,exporter MUST be
NULL
.