A globally unique identifier that combines ip, time, and random bits. Since the time is listed first, you can sort records by guid. You can also extract the time and ip if needed. GUIDs make wonderful database keys. They require no access to the database (to get the max index number), they are extremely unique, and they sort automatically by time.
Python, 183 lines
Download
Copy to clipboard1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 | #!/usr/bin/env python
# GUID.py
# Version 2.6
#
# Copyright (c) 2006 Conan C. Albrecht
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is furnished
# to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in all
# copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED,
# INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR
# PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE
# FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
# OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
# DEALINGS IN THE SOFTWARE.
##################################################################################################
### A globally-unique identifier made up of time and ip and 8 digits for a counter:
### each GUID is 40 characters wide
###
### A globally unique identifier that combines ip, time, and a counter. Since the
### time is listed first, you can sort records by guid. You can also extract the time
### and ip if needed.
###
### Since the counter has eight hex characters, you can create up to
### 0xffffffff (4294967295) GUIDs every millisecond. If your processor
### is somehow fast enough to create more than that in a millisecond (looking
### toward the future, of course), the function will wait until the next
### millisecond to return.
###
### GUIDs make wonderful database keys. They require no access to the
### database (to get the max index number), they are extremely unique, and they sort
### automatically by time. GUIDs prevent key clashes when merging
### two databases together, combining data, or generating keys in distributed
### systems.
###
### There is an Internet Draft for UUIDs, but this module does not implement it.
### If the draft catches on, perhaps I'll conform the module to it.
###
# Changelog
# Sometime, 1997 Created the Java version of GUID
# Went through many versions in Java
# Sometime, 2002 Created the Python version of GUID, mirroring the Java version
# November 24, 2003 Changed Python version to be more pythonic, took out object and made just a module
# December 2, 2003 Fixed duplicating GUIDs. Sometimes they duplicate if multiples are created
# in the same millisecond (it checks the last 100 GUIDs now and has a larger random part)
# December 9, 2003 Fixed MAX_RANDOM, which was going over sys.maxint
# June 12, 2004 Allowed a custom IP address to be sent in rather than always using the
# local IP address.
# November 4, 2005 Changed the random part to a counter variable. Now GUIDs are totally
# unique and more efficient, as long as they are created by only
# on runtime on a given machine. The counter part is after the time
# part so it sorts correctly.
# November 8, 2005 The counter variable now starts at a random long now and cycles
# around. This is in case two guids are created on the same
# machine at the same millisecond (by different processes). Even though
# it is possible the GUID can be created, this makes it highly unlikely
# since the counter will likely be different.
# November 11, 2005 Fixed a bug in the new IP getting algorithm. Also, use IPv6 range
# for IP when we make it up (when it's no accessible)
# November 21, 2005 Added better IP-finding code. It finds IP address better now.
# January 5, 2006 Fixed a small bug caused in old versions of python (random module use)
import math
import socket
import random
import sys
import time
import threading
#############################
### global module variables
#Makes a hex IP from a decimal dot-separated ip (eg: 127.0.0.1)
make_hexip = lambda ip: ''.join(["%04x" % long(i) for i in ip.split('.')]) # leave space for ip v6 (65K in each sub)
MAX_COUNTER = 0xfffffffe
counter = 0L
firstcounter = MAX_COUNTER
lasttime = 0
ip = ''
lock = threading.RLock()
try: # only need to get the IP addresss once
ip = socket.getaddrinfo(socket.gethostname(),0)[-1][-1][0]
hexip = make_hexip(ip)
except: # if we don't have an ip, default to someting in the 10.x.x.x private range
ip = '10'
rand = random.Random()
for i in range(3):
ip += '.' + str(rand.randrange(1, 0xffff)) # might as well use IPv6 range if we're making it up
hexip = make_hexip(ip)
#################################
### Public module functions
def generate(ip=None):
'''Generates a new guid. A guid is unique in space and time because it combines
the machine IP with the current time in milliseconds. Be careful about sending in
a specified IP address because the ip makes it unique in space. You could send in
the same IP address that is created on another machine.
'''
global counter, firstcounter, lasttime
lock.acquire() # can't generate two guids at the same time
try:
parts = []
# do we need to wait for the next millisecond (are we out of counters?)
now = long(time.time() * 1000)
while lasttime == now and counter == firstcounter:
time.sleep(.01)
now = long(time.time() * 1000)
# time part
parts.append("%016x" % now)
# counter part
if lasttime != now: # time to start counter over since we have a different millisecond
firstcounter = long(random.uniform(1, MAX_COUNTER)) # start at random position
counter = firstcounter
counter += 1
if counter > MAX_COUNTER:
counter = 0
lasttime = now
parts.append("%08x" % (counter))
# ip part
parts.append(hexip)
# put them all together
return ''.join(parts)
finally:
lock.release()
def extract_time(guid):
'''Extracts the time portion out of the guid and returns the
number of seconds since the epoch as a float'''
return float(long(guid[0:16], 16)) / 1000.0
def extract_counter(guid):
'''Extracts the counter from the guid (returns the bits in decimal)'''
return int(guid[16:24], 16)
def extract_ip(guid):
'''Extracts the ip portion out of the guid and returns it
as a string like 10.10.10.10'''
# there's probably a more elegant way to do this
thisip = []
for index in range(24, 40, 4):
thisip.append(str(int(guid[index: index + 4], 16)))
return '.'.join(thisip)
### TESTING OF GUID CLASS ###
if __name__ == "__main__":
guids = []
for i in range(10): # calculate very fast so people can see the counter in action
guid = generate()
guids.append(guid)
for guid in guids:
print "GUID:", guid
guidtime = extract_time(guid)
print "\tTime: ", time.strftime('%a, %d %b %Y %H:%M:%S', time.localtime(guidtime)), '(millis: ' + str(round(guidtime - long(guidtime), 3)) + ')'
print "\tIP: ", extract_ip(guid)
print "\tCounter:", extract_counter(guid)
|
This version uses a counter instead of random bits and is released under the MIT license. It is faster and should make for better GUIDs. I did not incorporate using a class as is done in the example below because I like GUIDs to remain simple strings. However, you could easily modify this new version to be a class if you prefer that method.
Tags: database
Very good but I would add format checking and __eq__... Here's a possibly improved version with basic GUID format checking, __eq__ implementation and a more pythonic ip implementation. Hope you like it.
thanks for sharing.
(comment continued...)
(...continued from previous comment)
(comment continued...)
(...continued from previous comment)
</pre>
Changes. Rodrigo -- Your code was cut off in the posting. Please send me the code to conan_albrecht@byu.edu and I'll integrate your changes back into the original and repost a new version. Thanks.
In fact, if others have more "pythonic" ways of doing things, let me know. I'm much more mature in other languages and relatively new to python.
Updated the code. I updated the code based upon reviews from many users. The main difference is the new GUID code simply generates a regular String object, rather than a GUID object. Why did I create a separate object? It seemed everyone was using GUIDs as strings anyway, so I removed the class and made functions instead.
The class now respects multiple threads, and it has been simplified somewhat.
int overflow under Python 2.2. Under Python 2.2 the phrase
fails with
while
succeeds.
Python 2.3 runs both successfully.
Maybe line 127
should better read
</pre>
Alternate version using descending sequence instead of random. Instead of random bits, I use a descending sequence. This uses a simple generator to start that segment at sys.maxint, decrementing by one on each call. 2.2 or later.
Another implementation with MIT-license is part of Pyro. For another GUID-generator, but licensed under a very liberal MIT software license, look in the Pyro.util package of Pyro. (http://pyro.sourceforge.net).
Random GUIDs. 16 bytes from a cryptographic-quality random number source like os.urandom() are just as good as a method of generating GUIDs.
Why I don't use the class version. Rodrigo -- nice class version. I just modified my original code recipe so yours needs to be updated if you like my new changes. The reason I don't use a class as you suggest is because potentially thousands of GUIDs are loaded from the DB at a time. Your class constructor has to parse the GUID every time it is created, which adds processing time. In addition, you then convert it back to a string to compare, which adds more time. Keeping it as a string doesn't require any additional processing time at create. Yours could actually be modified to compare the raw numbers instead of the string representation (which would be more efficient), but I still think it's easier to keep it as a string from the start. In the end it's just preference.
Recipe now uses counter. The recipe above now uses a variation of your method here. I use an increasing counter so its easier to sort them (why use a decreasing counter?), but otherwise I think this idea is now in the main recipe.
Best way to store in PostgreSql. Is it possible to convert the string to a number value and store in a NUMERIC type field in PostgreSql?
If you have pywin32.
the ip arument in the generate() method is never used. could you please fix it?
"GUID.py:92: FutureWarning: hex/oct constants > sys.maxint..." I get "GUID.py:92: FutureWarning: hex/oct constants > sys.maxint will return positive values in Python 2.4 and up" from Python 2.3.5 (which is pre-installed on Mac OS X 10.4), and the counter seems to always be 0!
I changed "MAX_COUNTER = 0xfffffffe" to "MAX_COUNTER = sys.maxint" which fixes the problem.
Use the uuid module instead. An industrial-strength uuid module is included in python 2.5+. 2.3 and 2.4 users can get the module from http://zesty.ca/python/uuid.html