Python:ctypes可散列的c_char数组替换，而不会导致'\ 0'字节跳闸 [英] Python: ctypes hashable c_char array replacement without tripping over '\0' bytes
问题描述
出于说明目的，此脚本创建一个文件mapfile，其中包含文件的内容(作为参数提供)，并以带有sha1校验和的二进制头开头，以允许在后续运行中进行重复检测.
For illustration purposes, this script creates a file mapfile containing the content of the files, given as arguments, prepended by a binary header with a sha1 checksum, that allows for duplication detection on subsequent runs.
这里需要的是可散列的ctypes.c_char替换，它可以以最小的模糊度保存sha1校验和，但不会阻塞'\ 0'字节.
What's needed here is a hashable ctypes.c_char replacement, that can hold sha1 checksums with minimum fuzz, but not choking on '\0' bytes.
# -*- coding: utf8 -*
import io
import mmap
import ctypes
import hashlib
import logging
from collections import OrderedDict
log = logging.getLogger(__file__)
def align(size, alignment):
"""return size aligned to alignment"""
excess = size % alignment
if excess:
size = size - excess + alignment
return size
class Header(ctypes.Structure):
Identifier = b'HEAD'
_fields_ = [
('id', ctypes.c_char * 4),
('hlen', ctypes.c_uint16),
('plen', ctypes.c_uint32),
('name', ctypes.c_char * 128),
('sha1', ctypes.c_char * 20),
]
HeaderSize = ctypes.sizeof(Header)
class CtsMap:
def __init__(self, ctcls, mm, offset = 0):
self.ctcls = ctcls
self.mm = mm
self.offset = offset
def __enter__(self):
mm = self.mm
offset = self.offset
ctsize = ctypes.sizeof(self.ctcls)
if offset + ctsize > mm.size():
newsize = align(offset + ctsize, mmap.PAGESIZE)
mm.resize(newsize)
self.ctinst = self.ctcls.from_buffer(mm, offset)
return self.ctinst
def __exit__(self, exc_type, exc_value, exc_traceback):
del self.ctinst
self.ctinst = None
class MapFile:
def __init__(self, filename):
try:
# try to create initial file
mapsize = mmap.PAGESIZE
self._fd = open(filename, 'x+b')
self._fd.write(b'\0' * mapsize)
except FileExistsError:
# file exists and is writable
self._fd = open(filename, 'r+b')
self._fd.seek(0, io.SEEK_END)
mapsize = self._fd.tell()
# mmap this file completely
self._fd.seek(0)
self._mm = mmap.mmap(self._fd.fileno(), mapsize)
self._offset = 0
self._toc = OrderedDict()
self.gen_toc()
def gen_toc(self):
while self._offset < self._mm.size():
with CtsMap(Header, self._mm, self._offset) as hd:
if hd.id == Header.Identifier and hd.hlen == HeaderSize:
self._toc[hd.sha1] = self._offset
log.debug('toc: [%s]%s: %s', len(hd.sha1), hd.sha1, self._offset)
self._offset += HeaderSize + hd.plen
else:
break
del hd
def add_data(self, datafile, data):
datasize = len(data)
sha1 = hashlib.sha1()
sha1.update(data)
digest = sha1.digest()
if digest in self._toc:
log.debug('add_data: %s added already', digest)
return None
log.debug('add_data: %s, %s bytes, %s', datafile, datasize, digest)
with CtsMap(Header, self._mm, self._offset) as hd:
hd.id = Header.Identifier
hd.hlen = HeaderSize
hd.plen = datasize
hd.name = datafile
hd.sha1 = digest
del hd
self._offset += HeaderSize
log.debug('add_data: %s', datasize)
blktype = ctypes.c_char * datasize
with CtsMap(blktype, self._mm, self._offset) as blk:
blk.raw = data
del blk
self._offset += datasize
return HeaderSize + datasize
def close(self):
self._mm.close()
self._fd.close()
if __name__ == '__main__':
import os
import sys
logconfig = dict(
level = logging.DEBUG,
format = '%(levelname)5s: %(message)s',
)
logging.basicConfig(**logconfig)
mf = MapFile('mapfile')
for datafile in sys.argv[1:]:
if os.path.isfile(datafile):
try:
data = open(datafile, 'rb').read()
except OSError:
continue
else:
mf.add_data(datafile.encode('utf-8'), data)
mf.close()
运行:python3 hashable_ctypes_bytes.py somefiles*
第二次调用它，它使用sha1摘要作为键，通读文件，将有序字典中的所有项目收集在一起.不幸的是，c_char数组的语义有点局限，因为它的也表现为类似于'\ 0'终止的c字符串，导致此处的校验和被截断.
Invoking it a second time, it reads through the file collecting all items in an ordered dict, using the sha1 digest as key. Unfortunately, the c_char array semantics are a little wired, because it also behaves like '\0' terminated c string, resulting in truncated checksums here.
请参见第3行和第4行:
See line 3 and 4:
DEBUG: toc: [20]b'\xcd0\xd7\xd3\xbf\x9f\xe1\xfe\xffr\xa6g#\xee\xf8\x84\xb5S,u': 0
DEBUG: toc: [20]b'\xe9\xfe\x1a;i\xcdG0\x84\x1b\r\x7f\xf9\x14\x868\xbdVl\x8d': 1273
DEBUG: toc: [19]b'\xa2\xdb\xff$&\xfe\x0f\xb4\xcaB<F\x92\xc0\xf1`(\x96N': 3642
DEBUG: toc: [15]b'O\x1b~c\x82\xeb)\x8f\xb5\x9c\x15\xd5e:\xa9': 4650
DEBUG: toc: [20]b'\x80\xe9\xbcF\x97\xdc\x93DG\x90\x19\x8c\xca\xfep\x05\xbdM\xfby': 13841
DEBUG: add_data: b'\xcd0\xd7\xd3\xbf\x9f\xe1\xfe\xffr\xa6g#\xee\xf8\x84\xb5S,u' added already
DEBUG: add_data: b'\xe9\xfe\x1a;i\xcdG0\x84\x1b\r\x7f\xf9\x14\x868\xbdVl\x8d' added already
DEBUG: add_data: b'../python/tmp/colorselect.py', 848 bytes, b'\xa2\xdb\xff$&\xfe\x0f\xb4\xcaB<F\x92\xc0\xf1`(\x96N\x00'
DEBUG: add_data: 848
DEBUG: add_data: b'../python/tmp/DemoCode.py', 9031 bytes, b'O\x1b~c\x82\xeb)\x8f\xb5\x9c\x15\xd5e:\xa9\x00p\x0f\xc04'
DEBUG: add_data: 9031
DEBUG: add_data: b'\x80\xe9\xbcF\x97\xdc\x93DG\x90\x19\x8c\xca\xfep\x05\xbdM\xfby' added already
通常的建议是将c_char * 20替换为c_byte * 20，并以此方式丢失透明字节的处理.除了数据转换麻烦之外，由于c_byte数组是字节数组，因此不可散列.我一直没有找到一种切实可行的解决方案，而无需一路来回经历繁重的转换麻烦，或者依靠十六进制，这会使sha1摘要大小要求加倍.
Usual suggestion is replace the c_char * 20 with c_byte * 20, and loosing the transparent bytes handling on that way. Apart from the data conversion hassles, c_byte arrays aren't hashable due to being bytearrays. I haven't found a practical solution without going through heavy conversion troubles all the way back and forth, or resorting to hexdigests, that doubles the sha1 digest size requirements.
我认为，将c_char与C零终止语义相混合的设计决定首先是错误的.为了解决这个问题，我可以想象在ctypes中添加一个c_char_nz类型，从而解决了这个问题.
I think, that the c_char design decision to mix it with C zero termination semantics was a fault in the first place. To cope with this, I could imagine to add a c_char_nz type to ctypes, that resolves this issue.
对于那些仔细阅读代码的人，您可能想知道ctypes结构的del语句.可以在此处找到它的讨论:.
For those of you, that read the code carefully, you might wonder about the del statements of the ctypes structures. A discussion of it can be found here:.
推荐答案
虽然下面的代码正在执行您提到的来回转换，但它确实很好地隐藏了问题.我发现一个包含空值的哈希，并且该字段现在可以用作字典键.希望对您有所帮助.
While the below code is doing the back-and-forth conversion you mentioned, it does nicely hide the issue. I found a hash that contained a null and the field can now be used as a dictionary key. Hope it helps.
from ctypes import *
import hashlib
class Test(Structure):
_fields_ = [('_sha1',c_ubyte * 20)]
@property
def sha1(self):
return bytes(self._sha1)
@sha1.setter
def sha1(self, value):
self._sha1 = (c_ubyte * 20)(*value)
test = Test()
test.sha1 = hashlib.sha1(b'aaaaaaaaaaa').digest()
D = {test.sha1:0}
print(D)
输出:
{b'u\\\x00\x1fJ\xe3\xc8\x84>ZP\xddj\xa2\xfa#\x89=\xd3\xad': 0}
这篇关于Python:ctypes可散列的c_char数组替换，而不会导致'\ 0'字节跳闸的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！
