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"x9dx01x00x00" 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：
