github的仓库是可以统计每个贡献者的代码行数的，公司年会的时候，特设了一个“码神奖”，颁给去年贡献代码最多的工程师，github的统计数据显示，这位大神去年提交的代码达到了110w行，这个数据太惊人了，一个人不可能写这么多代码的，我非常好奇的研究了一下，发现中间还包括了他提交的很多第三方库，但github也一并统计了，而且经过他合并的代码也会统计进去。那么有没有办法去掉这些无效数据，得到真实的代码贡献量呢？查了一下github api，再结合git 命令，还是可以的，上代码：

#copy this script to your target repo

#run python github-stats.py to collect data

import re

import json

import os

import sys

import requests

#get token from cmd line

tk = sys.argv[1]

user_stats={"dummy":{"additions":0,"deletions":0,"total":0}}

#query github api for last year's commits

payload = {'since':'2013-01-01T00:00:00Z','until':'2014-01-01T00:00:00Z','access_token':tk}

token = {'access_token':tk}

def is_merge(commit_sha):

cmd = "git show --oneline " + commit_sha

output = os.popen(cmd)

title = output.read()

p_merge = re.compile("Merge")

if(p_merge.search(title) is not None):

return True

else:

return False

def collect_stats(commit_list):

for m in commit_list:

#print user_stats

#print m['sha']

#print data

if(is_merge(m['sha'])):

continue

git_show_command = "git show -s --format=%an " + m['sha']

output = os.popen(git_show_command)

user = output.read().strip(' \t\n\r')

#print user

#r2 = requests.get(commit_request_api+m['sha'], params = token)

#commit = r2.json()

#print commit

git_diff_command = "git diff --shortstat "+m['sha'] + " " + m['sha'] + "^"

output = os.popen(git_diff_command)

data = output.read()

#print "data is:"

#print data

p_ins = re.compile("(\d+) insertion")

r_ins = p_ins.search(data)

ins_data = 0

del_data = 0

if(r_ins is not None):

ins_str = r_ins.group(1)

ins_data = int(ins_str)

#print ins_data

p_del = re.compile("(\d+) deletion")

r_del = p_del.search(data)

if(r_del is not None):

del_str = r_del.group(1)

del_data = int(del_str)

#print del_data

if(ins_data + del_data > 5000):

print user

print 'ins:'+str(ins_data)

print 'del:'+str(del_data)

ins_data = 0

del_data = 0

if(user in user_stats):

stats = user_stats[user]

stats['additions'] += ins_data

stats['deletions'] += del_data

stats['total'] += (ins_data + del_data)

user_stats[user] = stats

else:

new_stat = {'additions':ins_data, 'deletions':del_data, 'total':ins_data+del_data}

user_stats[user] = new_stat

r = requests.get("https://api.github.com/repos/cocos2d/cocos2d-x/commits", params = payload)

collect_stats(r.json())

print user_stats

pattern = re.compile("; rel=\"next\"")

h = r.headers

print r.headers['X-RateLimit-Remaining']

result = pattern.search(h['link'])

while(result is not None):

next_url = result.group(1)

r = requests.get(next_url, params = token)

collect_stats(r.json())

h = r.headers

print h['link']

result = pattern.search(h['link'])

#print h['link']

#next_url = result.group(1)

#print next_url

#r_next = requests.get(next_url[1])

print r.headers['X-RateLimit-Remaining']

print user_stats

代码也可以在github上获得: https://github.com/heliclei/githubtools/blob/master/github-stats.py

这个脚本过滤了单次提交超过5000行的commit，并且过滤了合并的commit，先把需要统计的仓库克隆到本地，再把这个脚本拷贝到本地git仓库下，注意要把这一行改为对应仓库的url

https://api.github.com/repos/cocos2d/cocos2d-x/commits

github token可以用上一篇的脚本生成

运行 python git-stats.py xxxxxxxxxxxxxgithub-oauth-tokenxxxxxxxxxxxxxxxxxxx

PS: 过滤后，cocos2d-x的码神去年的代码贡献量超过了10w行，还是非常的厉害～～但这个数据没有110W行那么超现实了。