Validation¶
(Production)
Timezone¶
Verify that``git``’s Unix timestamp match the current timezone.
$ git commit -a
$ date
dimanche 11 novembre 2018, 16:13:15 (UTC+0100)
$ git log --after "2018-11-11 16:00:00" --before "2018-11-11"
...
Date: Sun Nov 11 16:11:13 2018 +0100
$ ./keepcool.py -r . --after "2018-11-11 16:00:00" --before "2018-11-11" -v
...
Date(1541949073): Sunday 2018-11-11 16:11:13 -> ['week-end']
Branches¶
Check on a repository having several branches.
$ mkdir linux
$ cd linux
$ git clone https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/
$ git branch
* master
No branches on linux sources, I guess I miss something...
(but it seems to me that git log returns by default the commits for all branches.)
Performances¶
Use a profiler tool on a udge (800.000 commits) repository.
$ ./keepcool.py linux/ -s 'sum'
...
Mauro Carvalho Chehab: 29462 / 48%
Greg Kroah-Hartman: 74367 / 70%
Linus Torvalds: 75151 / 68%
David S. Miller: 77947 / 74%
-------------------------------------
all: 796965 / 54%
-------------------------------------
$ python -m cProfile -o linux.profile keepcool.py linux
$ sudo pip install cprofilev
$ cprofilev -f linux.profile
cProfileV]: cProfile output available at http://127.0.0.1:4000
13494147 function calls (13492172 primitive calls) in 19.650 seconds
Ordered by: internal time
ncalls tottime percall cumtime percall filename:lineno(function)
746797 8.571 0.000 8.571 0.000 {built-in method poll}
796965 2.025 0.000 5.867 0.000 keepcool.py:252(add_commit)
796965 1.392 0.000 1.392 0.000 {built-in method fromtimestamp}
796965 1.053 0.000 2.446 0.000 keepcool.py:27(__init__)
1 0.907 0.907 17.516 17.516 keepcool.py:190(parse_logs)
1 0.693 0.693 9.933 9.933 /usr/lib/python2.7/subprocess.py:1122(_communicate_with_poll)
746798 0.621 0.000 0.621 0.000 {posix.read}
796965 0.521 0.000 1.565 0.000 keepcool.py:74(belongs_to_workin_hours)
796965 0.499 0.000 2.945 0.000 keepcool.py:92(__init__)
...
796965 0.290 0.000 0.290 0.000 keepcool.py:230(get_user)
We can see that the longest times spent into functions (add_commit...), are only seven time bigger than a unique dictionary access (get_user). By the way, we should better optimise the model than the code in order to be quicker.
Memory footprint¶
Check the memory usage while running.
$ while /bin/true; do sleep 2; ps -e -ovsz -orss,args= | grep [k]eepcool; done
25184 20196 python ./keepcool.py linux
40384 35372 python ./keepcool.py linux
55532 50536 python ./keepcool.py linux
70876 65832 python ./keepcool.py linux
87572 82612 python ./keepcool.py linux
533500 528136 python ./keepcool.py linux
653648 648432 python ./keepcool.py linux
771196 766052 python ./keepcool.py linux
376008 371328 python ./keepcool.py linux
$ while /bin/true; do sleep 2; ps -e -ovsz -orss,args= | grep [g]it; done
233292 180476 git log --encoding=UTF-8 --pretty=%H; %cN; %ct
349572 269532 git log --encoding=UTF-8 --pretty=%H; %cN; %ct
434652 356148 git log --encoding=UTF-8 --pretty=%H; %cN; %ct
500564 410912 git log --encoding=UTF-8 --pretty=%H; %cN; %ct
524036 462844 git log --encoding=UTF-8 --pretty=%H; %cN; %ct
The script does not use too much memory compare to the git log query it launches.