Question

我正在尝试使用Python中zlib的crc32_combine函数。虽然有各种其他zlib功能可用，这个不是“包含电池”标准库的一部分。我尝试了两种方法：从C代码到Python的端口用ctypes从Python调用zlib。两者都给我不同的结果，虽然不是我期待的结果。我提出了ctypes代码，因为我认为这样执行得更快并且有一个其他程序员错误的可能性较小。

算法可以combine two CRC32哈希值时的长度提供第二散列的数据。 crc32_combine定义如下：

crc32(crc32(0, seq1, len1), seq2, len2) == crc32_combine(
    crc32(0, seq1, len1), crc32(0, seq2, len2), len2)

这是输出：

Expected CRC: 45E57586
Combined CRC: 567EE4E4

在win32上使用Python 3.5.1运行时，第二行总是不同的。不是Python 2，但结果永远不是我所期望的。将zlib1.dll放在与脚本相同的目录中以进行试用。

import zlib

def crc32_combine_ctypes(crc1, crc2, len2):
    import ctypes
    from ctypes import util

    lib = util.find_library('zlib1')
    _zlib = ctypes.CDLL(lib)
    assert _zlib._name, "Can't find zlib"

    _zlib.crc32_combine.argtypes = [
        ctypes.c_ulong, ctypes.c_ulong, ctypes.c_ulong]
    _zlib.crc32_combine.restype = ctypes.c_ulong

    return _zlib.crc32_combine(crc1, crc2, len2)

testfile = "zlib1.dll"

with open(testfile, "rb") as tf:
    data = tf.read()

print("Expected CRC: %0.8X" % (zlib.crc32(data) & 0xFFFFFFFF))

cut = len(data) // 2 - 100
p1 = data[0:cut]
p2 = data[cut:]

crc1 = zlib.crc32(p1)
crc2 = zlib.crc32(p2)
len1 = len(p1)
len2 = len(p2)

combined = crc32_combine_ctypes(crc1, crc2, len2)
print("Combined CRC: %0.8X" % (combined & 0xFFFFFFFF))

我做错了什么？

Answer 1

eryksun有正确的想法：我使用了一个糟糕的旧DLL。包含32位dll的最后一个zlib版本包括： https://sourceforge.net/projects/libpng/files/zlib/1.2.8/

我的纯Python代码端口比使用ctypes调用库慢几百倍。（使用带有1k次迭代和50m作为长度参数的timeit的数字）

31.729 (function provided below)
 0.120 (just the _zlib.crc32_combine() call: no library loading included)

纯Python代码：

def crc32_combine(crc1, crc2, len2):
    """Explanation algorithm: https://stackoverflow.com/a/23126768/654160
    crc32(crc32(0, seq1, len1), seq2, len2) == crc32_combine(
        crc32(0, seq1, len1), crc32(0, seq2, len2), len2)"""
    # degenerate case (also disallow negative lengths)
    if len2 <= 0:
        return crc1

    # put operator for one zero bit in odd
    # CRC-32 polynomial, 1, 2, 4, 8, ..., 1073741824
    odd = [0xedb88320] + [1 << i for i in range(0, 31)]
    even = [0] * 32

    def matrix_times(matrix, vector):
        number_sum = 0
        matrix_index = 0
        while vector != 0:
            if vector & 1:
                number_sum ^= matrix[matrix_index]
            vector = vector >> 1 & 0x7FFFFFFF
            matrix_index += 1
        return number_sum

    # put operator for two zero bits in even - gf2_matrix_square(even, odd)
    even[:] = [matrix_times(odd, odd[n]) for n in range(0, 32)]

    # put operator for four zero bits in odd
    odd[:] = [matrix_times(even, even[n]) for n in range(0, 32)]

    # apply len2 zeros to crc1 (first square will put the operator for one
    # zero byte, eight zero bits, in even)
    while len2 != 0:
        # apply zeros operator for this bit of len2
        even[:] = [matrix_times(odd, odd[n]) for n in range(0, 32)]
        if len2 & 1:
            crc1 = matrix_times(even, crc1)
        len2 >>= 1

        # if no more bits set, then done
        if len2 == 0:
            break

        # another iteration of the loop with odd and even swapped
        odd[:] = [matrix_times(even, even[n]) for n in range(0, 32)]
        if len2 & 1:
            crc1 = matrix_times(odd, crc1)
        len2 >>= 1

        # if no more bits set, then done
    # return combined crc
    crc1 ^= crc2
    return crc1

在Python

1 个答案: