生成全局唯一Id
来源:互联网 发布:易语言充值源码 编辑:程序博客网 时间:2024/05/01 01:29
生成全局唯一Id
参考了: http://www.cnblogs.com/heyuquan/p/global-guid-identity-maxId.html
- GUID
优点: 确保唯一, 速度快
缺点: 太长, 不友好, 不好索引
- 数据库唯一索引
时间戳加上随机数,然后通过数据库做唯一性校验
import timeimport randomimport stringm = time.strftime('%y%m%d%H%M%S') + ''.join([random.choice(string.lowercase + string.digits) for _ in range(5)])#检查m在数据库中是否存在,存在则重复上述过程,不存在则存入数据库并返回
优点:适合简单应用,id较短,有一定亲和力
缺点:每秒id总数有限制,并发越大性能越低, 加大数据库访问压力,需要锁表
优化:将时间戳转成62进制数
digit62 = '0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz'#整数转化为62进制字符串#入口:# x : 整数#返回: 字符串def int_to_str62( x ): try: x=int(x) except: x=0 if x<0: x=-x if x==0: return "0" s="" while x>=62: x1= x % 62 s = digit62[x1]+s x = x // 62 if x>0: s = digit62[x]+s return s#62进制字符串转化为整数#入口:# s : 62进制字符串#返回: 整数def str62_to_int( s ): x = 0 s = str(s).strip() if s=="": return x for y in s: k = digit62.find(y) if k>=0: x=x*62+k return ximport timeimport randomimport stringt = time.strftime('%y%m%d%H%M%S')cut = [ t[i:i+2] for i in range(0, len(t), 2) ]62t = ''.join([ int_to_str62(int(x)) for x in cut])m = 62t + ''.join([random.choice(string.lowercase + string.digits) for _ in range(6)])
再ps. 有人说random.choice慢而且随机不均匀,我就写了两个小程序测试一下
import randomimport timeimport stringimport timeitimport hashlibimport uuidimport threadingdef randomchoice(): return ''.join([ random.choice(string.lowercase + string.digits) for _ in range(6)])def _time(f, n=1000000): print 'start timeit function ', f t = timeit.timeit(f, number=n) print 'repeat %s times and used %ss' % (n, t) print 'end timeit function ', f print_time(randomchoice)
result
start timeit function <function randomchoice at 0x2a7d6e0>repeat 1000000 times and used 3.97338795662send timeit function <function randomchoice at 0x2a7d6e0>
随机分布
from random import choiceimport stringimport collectionsfrom matplotlib.pyplot import plot, show, barh, yticks, xlabel, title, figureimport numpy as nptables = string.ascii_letters + string.digitscounter = collections.Counter()for _ in range(1000000): counter[choice(tables)] += 1alphats = counter.keys()y_pos = np.arange(len(alphats))freq = counter.values()figure(figsize=(100,100))barh(y_pos, freq, align='edge', alpha=1, height=0.05)yticks(y_pos, alphats)xlabel('frequence')title('random choice')show()
结果图:
可见分布还是比较平均的
- like mongo objectid
时间 + md5(hostname) + pid + 递增id
import structimport socketimport osimport timefrom hashlib import md5import threadingimport randomimport binascii_inc = random.randint(0, 0xFFFFFF)_inc_lock = threading.Lock()oid = ""oid += struct.pack(">i", int(time.time()))m = md5()m.update(socket.gethostname())oid += m.digest()[0:3]oid += struct.pack(">H", os.getpid() % 0xFFFF)_inc_lock.acquire()oid += struct.pack(">i", _inc)[1:4]_inc = (_inc + 1) % 0xFFFFFF_inc_lock.release()print len(oid)print binascii.hexlify(oid)