[LLVMdev] "surprising" empirical numbers about git vs. svn

[Sean Silva]

I just ran some real, hard numbers about git and svn usage. These are
just numbers, I don't want people to jump to conclusions.

I basically just wrote a little script that iterates over every
message in llvm-commits (for as long as I have been subscribed, which
is about since Feb 2012) and if there are patches attached it tallies
up the counts of what kind of patch it looks like.

It uses extremely simple (but apparently effective) heuristics to
determine whether a patch is from svn or git. The script is attached
if you want to take a peek.

The results are as follows:

sean:~/pg/llvm/patchcount % ./patchcount.py
error    2
git      1082
svn      1057
unknown  23
vanilla  28

So basically, git and svn are roughly the same for patches that are
received on llvm-commits. Obviously, this ignores a number of things
that are relevant to determining the prevalence of svn vs git for
*general development*. These numbers indicate exactly what it says on
the wrapper: number of patches sent to llvm-commits that are generated
by each program.

-- Sean Silva
-------------- next part --------------
A non-text attachment was scrubbed...
Name: patchcount.py
Type: application/octet-stream
Size: 1791 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20121130/cd46dcff/attachment.obj>