问题
i have a Question about Pythons ctypes and calling C functions bothering me for a couple days now. I'm working with Python 3.5 and ctypes to wrap a C .dll.
I've got a C Function accepting a void ** as an output parameter. This should contain a Pointer to some 8 Bit RGB Data of a Image after the Call.
foo(void ** bar)
I declare my Python parameter and function call as following:
>>bar = c_void_p()
>>foo(byref(bar))
And try to recieve the Data:
>>data = c_byte(bar.value)
>>data.value
20
which actually is a reasonable value for the first Pixels 'R'-Byte But i did not manage to reach any Byte beyond this. For my understandig I now get the value of the first Byte the c_void_p points to(?)
I also tried:
>>data = (c_byte*3)(bar.value)
3 is only for test purposes (tried and went up to couple thousands):
>>data
<__main__.c_byte_Array_3 at 0x7faca45de950> #so far so good again?
>>data[0]
20
>>data[1]
0
...
>>data[n]
0
for me it seems like i cant acces any further Bytes again?
when i Try to define bar as a int Pointer (what is recommend by a lot of guys) i get the following:
>>bar = POINTER(c_int)()
>>foo(byref(bar))
>>bar
<__main__.LP_c_int at 0x7fcf71d9d8c8> #so far so good?
>>bar.contents
Process finished with exit code 139 (interrupted by signal 11: SIGSEGV)
I also worked through anything i found online including: this, this, and a couple more I cant find right now.
Any Help is really appreciated!
edit: whole code of the python file:
from ctypes import *
from ctypes_wrapper_dijsdk import *
init()
guids = find_cameras()
handle = c_void_p()
open_camera(guids, handle, 0)
start_acquisition(handle, 1)
imageHandle = c_void_p()
VOIDPP = POINTER(c_void_p)
imageData = VOIDPP(c_void_p())
get_image(handle, imageHandle, imageData)
byte_buf = cast(imageData.contents, POINTER(c_ubyte * (1920)))
for x in range(0, (1920)-1 ):
print("value: {:3d}".format(byte_buf.contents[x]))
dimension = (c_int*2)()
get_int_parameter(imageHandle, 0x20000103, dimension, 2)
value = c_void_p()
get_int_parameter(imageHandle, 0x20000200, value)
#from now on would be able to get the ImageData Size
print("ImageFormat = ", value.value, "see SDK.h")
release_image(imageHandle)
close_camera(handle)
exit()
call from python to C:
def get_image(handle, imageHandle, imageData, timeout = 0):
return d.SDK_GetImage(handle, byref(imageHandle), imageData, timeout)
additional information:
there actually are more
void**output parameters in the code, such asimageHandle. But they do not address any Arrays and seem to work just fine. They are declared asc_void_pand are calledbyref( ).the RGB Data stored as a Byte stream of Pixels of each Line. so its dimension[0] x dimension[1] x bit-depth long.
I can't provide any C Code because it is part of a shared library.
回答1:
This is one way of doing things. Also, ctypes's official doc: [Python 3.5]: ctypes - A foreign function library for Python.
dll.c:
#include <stdio.h>
#include <stdlib.h>
#if defined(_WIN32)
# define DLL_EXPORT __declspec(dllexport)
#else
# define DLL_EXPORT
#endif
#define C_TAG "From C"
#define PRINT_MSG_0() printf("%s - [%s] (%d) - [%s]\n", C_TAG, __FILE__, __LINE__, __FUNCTION__)
#define PRINT_ERR_1S(ARG0) printf("%s: %s\n", C_TAG, ARG0)
DLL_EXPORT int test(void **pptr, size_t count) {
PRINT_MSG_0();
if (!pptr) {
PRINT_ERR_1S("NULL pointer received");
return -1;
}
if (*pptr) {
PRINT_ERR_1S("Non NULL inner pointer received");
return -2;
}
unsigned char *buf = (unsigned char*)malloc(count);
for (size_t i = 0; i < count; i++) {
buf[i] = (i + 1) % 0x100;
}
*pptr = buf;
return 0;
}
DLL_EXPORT void dealloc(void **pptr) {
PRINT_MSG_0();
if ((pptr) && (*pptr)) {
free(*pptr);
*pptr = NULL;
}
}
code.py:
import sys
import math
from ctypes import c_ubyte, c_int, c_size_t, c_void_p, \
POINTER, CDLL, \
cast
VoidPtrPtr = POINTER(c_void_p)
dll_dll = CDLL("./dll.dll")
test_func = dll_dll.test
test_func.argtypes = [VoidPtrPtr, c_size_t]
test_func.restype = c_int
dealloc_func = dll_dll.dealloc
dealloc_func.argtypes = [c_void_p]
DISPLAY_VALUES_COUNT = 5
FORMAT_STRING_PAT = " idx: {{:{:d}d}} - value: {{:3d}}"
def _get_print_indexes(array_size, values_count):
if array_size <= 0 or values_count <= 1 or values_count > array_size:
raise ValueError("Invalid args")
yield 0
if array_size > 1:
if values_count > 2:
interval_size = array_size / (values_count - 1)
for idx in range(1, values_count - 1):
yield int(round(idx * interval_size))
yield array_size - 1
def _print_array_values(array, array_size, values_count=DISPLAY_VALUES_COUNT):
index_width = math.ceil(math.log10(array_size))
format_string = FORMAT_STRING_PAT.format(index_width)
for idx in _get_print_indexes(array_size, values_count):
print(format_string.format(idx, array.contents.contents[idx]))
def main():
sizes = [
10,
100,
500,
1920 * 1080 * 3,
]
for size in sizes:
UByteArr = c_ubyte * size
UByteArrPtr = POINTER(UByteArr)
UByteArrPtrPtr = POINTER(UByteArrPtr)
print("\nSize: {:d}".format(size))
data = UByteArrPtrPtr(UByteArrPtr())
print("data: {:}, data.contents: {:}".format(data, data.contents))
#print(addressof(data), addressof(data.contents))
ptr = cast(data, VoidPtrPtr)
res = test_func(ptr, size)
if res < 0:
print("{:s} returned {:d}. Moving on...\n".format(test_func.__name__, res))
continue
print("data: {:}, data.contents: {:}".format(data, data.contents))
_print_array_values(data, size)
dealloc_func(data)
print("data: {:}, data.contents: {:}".format(data, data.contents))
if __name__ == "__main__":
print("Python {:s} on {:s}\n".format(sys.version, sys.platform))
main()
Notes:
- Since the array allocation place (C or Python) wasn't mentioned, I chose the former option (otherwise, the double pointer (
void **) wouldn't make much sense). This adds some complexity to the code (among other things:dealloc- to avoid memory leaks). Using the latter option would "generate" less code, and no need forvoid**(I'm still wondering why it's needed anyway) void*is a generic type with not very much info. That's why in C, in order to populate the individual bytes, I had to cast it tounsigned char *. Same thing applies to Python. The way I expressedvoid **in Python is viaPOINTER(c_void_p)- A lot of the code is for printing purposes. That includes:
_get_print_indexes, which is only used to select 5 (DISPLAY_VALUES_COUNT) equidistant (from index perspective) elements in an array_print_array_values- prints the values
Output:
(py35x64_test) e:\Work\Dev\StackOverflow\q051981858>"c:\Install\x86\Microsoft\Visual Studio Community\2015\vc\vcvarsall.bat" x64 (py35x64_test) e:\Work\Dev\StackOverflow\q051981858>dir /b code.py dll.c (py35x64_test) e:\Work\Dev\StackOverflow\q051981858>cl /nologo dll.c /DDLL /link /DLL /OUT:dll.dll dll.c Creating library dll.lib and object dll.exp (py35x64_test) e:\Work\Dev\StackOverflow\q051981858>dir /b code.py dll.c dll.dll dll.exp dll.lib dll.obj (py35x64_test) e:\Work\Dev\StackOverflow\q051981858>"e:\Work\Dev\VEnvs\py35x64_test\Scripts\python.exe" code.py Python 3.5.4 (v3.5.4:3f56838, Aug 8 2017, 02:17:05) [MSC v.1900 64 bit (AMD64)] on win32 Size: 10 data: <__main__.LP_LP_c_ubyte_Array_10 object at 0x000001F8189FBBC8>, data.contents: <__main__.LP_c_ubyte_Array_10 object at 0x000001F8189FBC48> From C - [dll.c] (16) - [test] data: <__main__.LP_LP_c_ubyte_Array_10 object at 0x000001F8189FBBC8>, data.contents: <__main__.LP_c_ubyte_Array_10 object at 0x000001F8189FBCC8> idx: 0 - value: 1 idx: 2 - value: 3 idx: 5 - value: 6 idx: 8 - value: 9 idx: 9 - value: 10 From C - [dll.c] (35) - [dealloc] data: <__main__.LP_LP_c_ubyte_Array_10 object at 0x000001F8189FBBC8>, data.contents: <__main__.LP_c_ubyte_Array_10 object at 0x000001F8189FBD48> Size: 100 data: <__main__.LP_LP_c_ubyte_Array_100 object at 0x000001F8189FBCC8>, data.contents: <__main__.LP_c_ubyte_Array_100 object at 0x000001F8189FBDC8> From C - [dll.c] (16) - [test] data: <__main__.LP_LP_c_ubyte_Array_100 object at 0x000001F8189FBCC8>, data.contents: <__main__.LP_c_ubyte_Array_100 object at 0x000001F8189FBC48> idx: 0 - value: 1 idx: 25 - value: 26 idx: 50 - value: 51 idx: 75 - value: 76 idx: 99 - value: 100 From C - [dll.c] (35) - [dealloc] data: <__main__.LP_LP_c_ubyte_Array_100 object at 0x000001F8189FBCC8>, data.contents: <__main__.LP_c_ubyte_Array_100 object at 0x000001F8189FBE48> Size: 500 data: <__main__.LP_LP_c_ubyte_Array_500 object at 0x000001F8189FBC48>, data.contents: <__main__.LP_c_ubyte_Array_500 object at 0x000001F8189FBEC8> From C - [dll.c] (16) - [test] data: <__main__.LP_LP_c_ubyte_Array_500 object at 0x000001F8189FBC48>, data.contents: <__main__.LP_c_ubyte_Array_500 object at 0x000001F8189FBDC8> idx: 0 - value: 1 idx: 125 - value: 126 idx: 250 - value: 251 idx: 375 - value: 120 idx: 499 - value: 244 From C - [dll.c] (35) - [dealloc] data: <__main__.LP_LP_c_ubyte_Array_500 object at 0x000001F8189FBC48>, data.contents: <__main__.LP_c_ubyte_Array_500 object at 0x000001F8189FBF48> Size: 6220800 data: <__main__.LP_LP_c_ubyte_Array_6220800 object at 0x000001F8189FBDC8>, data.contents: <__main__.LP_c_ubyte_Array_6220800 object at 0x000001F818A62048> From C - [dll.c] (16) - [test] data: <__main__.LP_LP_c_ubyte_Array_6220800 object at 0x000001F8189FBDC8>, data.contents: <__main__.LP_c_ubyte_Array_6220800 object at 0x000001F8189FBEC8> idx: 0 - value: 1 idx: 1555200 - value: 1 idx: 3110400 - value: 1 idx: 4665600 - value: 1 idx: 6220799 - value: 0 From C - [dll.c] (35) - [dealloc] data: <__main__.LP_LP_c_ubyte_Array_6220800 object at 0x000001F8189FBDC8>, data.contents: <__main__.LP_c_ubyte_Array_6220800 object at 0x000001F8189FBEC8>
@EDIT0:
- Cleaned the code a bit (some renames, ...)
- Separated printing (added
_print_array_values) - Modified the ctypes pointers to be more meaningful. What stands out and can be confusing, is the relationship between a pointer and an array (when wrapped by another (outer) pointer):
- In C they are somewhat equivalent: both reference the address of the 1st (array) element. Easily cast one to another
- In Python the array must be wrapped by
POINTER(like a pass by value type) in order to be equivalent to the pointer (and cast from / to it)
- Successfully ran the same code on Lnx
来源：https://stackoverflow.com/questions/51981858/dereferencing-the-whole-data-of-c-void-p-not-only-the-first-byte