本来想写一个机器人成语接龙的，太复杂了，弄一个可以找到所有的可以“一招制敌”的成语，就是没办法再被接龙的成语。

第一步是找一个带拼音的成语字库，网上找了半天，没找到合适的，发现搜狗输入法自带成语词库，就拿来解析了。

搜狗的字库是scel格式，内容包括拼音表和字库两部分，它们的结构如下所示：

解析代码如下 [parse_scel.py]：为了避开烦人的字符编码问题，用的是python 3

增加了print，方便跟踪结果# encoding=utf8

# python 3

import sysimport structPIN_YIN_TAG_LEN = 4

def get_hanzi_offset(buff):

mask = buff[4]

return 0x2628 if mask == 0x44 else 0x26c4 if mask == 0x45 else None

def get_pinyin_offset(buff):

offset = 0x1540

return offset if buff[offset:offset + PIN_YIN_TAG_LEN] == b"\x9d\x01\x00\x00" else None

def get_piyin_table(buff, py_offset, hz_offset):

py_table = {}

data = buff[py_offset + PIN_YIN_TAG_LEN:hz_offset] pos = 0

while pos < len(data):

# 序号

index = struct.unpack('H', data[pos:pos + 2])[0]

pos += 2

# 拼音长度

l = struct.unpack('H', data[pos:pos + 2])[0]

pos += 2

# 拼音

py = data[pos:pos + l]

pos += l

py_table[index] = py.decode('UTF-16LE')

print(index, l, py.decode('UTF-16LE'))

return py_table

def get_hanzi_and_pinyin(buff, hz_offset, py_table): def _get_py(data, py_table):

length = len(data)

pos = 0

py = u''

while pos < length:

index = struct.unpack('H', data[pos:pos + 2])[0]

py += u'.' + py_table[index]

pos += 2

return py[1:] hz_all = {}

pos = 0

hz_table = buff[hz_offset:]

while pos < len(hz_table):

# 同音词数量

same = struct.unpack('H', hz_table[pos:pos + 2])[0]

pos += 2

# 拼音索引表长度

py_table_len = struct.unpack('H', hz_table[pos:pos + 2])[0]

pos += 2

# 拼音索引表

data = hz_table[pos:pos + py_table_len]

py = _get_py(data, py_table)

pos += py_table_len print(same, py_table_len, py)

# 中文词组

for i in range(same):

# 中文词组长度

hz_len = struct.unpack('H', hz_table[pos:pos + 2])[0]

pos += 2

# 中文词组

word = hz_table[pos:pos + hz_len]

pos += hz_len

# 扩展数据长度

ext_len = struct.unpack('H', hz_table[pos:pos + 2])[0]

pos += 2

# 序号

count = struct.unpack('H', hz_table[pos:pos + 2])[0]

pos += ext_len

hz_all[count] = (word.decode('UTF-16LE'), py)

print(hz_len, word.decode('UTF-16LE'), ext_len, count)

return hz_all

def scel_to_txt(scel, out):

with open(scel, "rb") as src, open(out, "w") as dst:

buff = src.read()

hz_offset = get_hanzi_offset(buff)

py_offset = get_pinyin_offset(buff)

if not hz_offset or not py_offset:

print("scel format changed, not support now \

[hz_offset = {} py_offset = {}]!".format(hz_offset, py_offset))

sys.exit(1) py_table = get_piyin_table(buff, py_offset, hz_offset)

hz_py_all = get_hanzi_and_pinyin(buff, hz_offset, py_table)

for x in hz_py_all:

dst.write(hz_py_all[x][0] + ' ' + hz_py_all[x][1] + '\n')

pass

if __name__ == "__main__":

scel_to_txt(u"成语大全.scel", u"成语大全.txt")

pass

生成的结果截图如下：