homepage

I'm trying to create ctypes buffer to be used back and forth with C libraries, but I keep getting errors. I need to slice the buffer to cut out different pieces to work on, so I try to get a memoryview of the buffer. does the error message imply that c_ubyte, c_uint8, c_byte and c_char are not native single character types?
>>> memoryview(c_ubyte()).cast('B')
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
ValueError: memoryview: source format must be a native single character format prefixed with an optional '@'
>>> memoryview((c_ubyte*1024)()).cast('B')
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
ValueError: memoryview: source format must be a native single character format prefixed with an optional '@'
>>> memoryview(c_uint8()).cast('B')
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
ValueError: memoryview: source format must be a native single character format prefixed with an optional '@'
>>> memoryview((c_uint8*1024)()).cast('B')
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
ValueError: memoryview: source format must be a native single character format prefixed with an optional '@'
>>> memoryview((c_byte*1024)()).cast('B')
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
ValueError: memoryview: source format must be a native single character format prefixed with an optional '@'
>>> memoryview((c_char*1024)()).cast('B')
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
ValueError: memoryview: source format must be a native single character format prefixed with an optional '@'
>>> memoryview((c_char*1024)())[0]
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
NotImplementedError: memoryview: unsupported format (1024)<c
>>> memoryview((c_byte*1024)())[0]
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
NotImplementedError: memoryview: unsupported format (1024)<b
>>> memoryview((c_ubyte*1024)())[0]
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
NotImplementedError: memoryview: unsupported format (1024)<B
>>> memoryview((c_uint8*1024)())[0]
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
NotImplementedError: memoryview: unsupported format (1024)<B
>>>

Memoryview currently only knows the types from the struct module.

this seems to disagree with the statement of "Memoryview currently only knows the types from the struct module."
why is memoryview aware of _pack_, a implementation detail of ctypes structures. is memoryview a collection of implementation for different types or a unique type on its own?
>>> import ctypes
>>> class B1(ctypes.Structure):
... _fields_ = [( "data", c_uint8 * 256 ), ]
... _pack_ = 1
...
>>> class B2(ctypes.Structure):
... _fields_ = [( "data", c_uint8 * 256 ), ]
...
>>>
>>> a = B1()
>>> b = B2()
>>> memoryview( a ).cast( 'B' )
<memory at 0x01FFCDC0>
>>> memoryview( b ).cast( 'B' )
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
ValueError: memoryview: source format must be a native single character format prefixed with an optional '@'
>>>

more examples (using 64-bit integer vs 8-bit integer in the above example) to show that ctypes aren't being translated for memoryview properly. _pack_ is the only way to make memoryview handle ctypes properly
>>> import ctypes
>>> class B1(ctypes.Structure):
... _fields_ = [( "data", ctypes.c_uint64 * 256 ), ]
... _pack_ = 1
...
>>> class B2(ctypes.Structure):
... _fields_ = [( "data", ctypes.c_uint64 * 256 ), ]
...
>>>
>>> a = B1()
>>> b = B2()
>>> memoryview( a ).cast( 'B' )
<memory at 0x01FFB030>
>>> memoryview( b ).cast( 'B' )
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
ValueError: memoryview: source format must be a native single character format prefixed with an optional '@'
>>>

>>> class B1(ctypes.Structure):
... _fields_ = [("data", ctypes.c_uint8 * 256), ]
... _pack_ = 1 
... 
>>> a= B1()
>>> x = memoryview(a)
>>> x.format
'B'
In the first case the format is 'B', in the second case the
format is:
>>> x = memoryview(b)
>>> x.format
'T{(256)<B:data:}'
While the latter is probably a valid PEP 3118 format, it's
not implemented anywhere outside ctypes See #3132.

I wonder why "_pack_ = 1" has significant impact on the behaviour of memoryview if memoryview is not a sub class of ctypes structure. unless memoryview was specifically coded to understand ctypes structure, I don't see why "_pack_ = 1" should make any difference.

Memoryview sends a *request* to ctypes to fill in a Py_buffer struct according
to the buffer *protocol*. IOW, memoryview knows nothing about ctypes.
For some reason ctypes fills in 'B' as the format if _pack_ is set. I do not
know if that is intended, it could be a ctypes issue.