[Qgis-developer] size of git repository

[Sandro Santilli]

Hello everybody,
today, after being waiting too much on a "git log --stat" command I went
asking for ideas in the #git channel, and reported the QGIS git repo url.

They cloned and reaction was like this:

 12:42 < luto> strk: what is in there? an ubuntu iso? :D
 12:44 < canton7> 925 megs in all

I went digging up history back to June 2011, when the code was moved
to git, and according to my blog post at that time the size was 330 Mb:
http://strk.keybit.net/blog/2011/06/07/getting-just-the-tip-of-a-remote-git-branch/

 12:48 < luto> 200mb of XML
 12:49 < luto> 4mb per file
 12:50 < canton7> yeah, those xml files are like 60000 lines each
 12:51 < canton7> strk, in conclusion, I'm not surprised --stat's being slow.
                  calculating which 50 lines have changed out of 60000
                  for a few files at a time is going to take a little
                  while :P
 12:51 < canton7> 1.3 gigs for a whole clone

There might be nothing that can be done to reduce the size of the
repository w/out altering the history, but maybe some care should be put
into avoiding to make the growth so fast.

 12:56 < luto> $ time git --no-pager log --stat > /dev/null
 12:56 < luto> 167.40s user 2.13s system 100% cpu 2:49.38 total
 12:56 < luto> $ time git --no-pager log > /dev/null
 12:56 < luto> 0.34s user 0.02s system 99% cpu 0.357 total
 12:56 < luto> on a ramfs and i7-3630QM (quad core, 2.4ghz)
 12:56 < luto> that's really not something one can work with..

On my laptop it takes 4 minutes to run this:

  time git --no-pager log --stat HEAD~100 > /dev/null

In the #git channel on freenode are going on with the analisys to try
at figure out if it's possible to alleviate the speed problem.
I though it was worth mentioning to the list.

--strk;

  ()   Free GIS & Flash consultant/developer
  /\   http://strk.keybit.net/services.html