Russian article about Bazaar

Alexander Belchenko bialix at ukr.net
Thu Nov 1 06:59:26 GMT 2007


Robert Collins пишет:
> On Tue, 2007-10-30 at 20:29 +0200, Alexander Belchenko wrote:
>> There are also commentaries from different ukranian and russian
>> developers,
>> but the tone is mostly negative about Bazaar speed.
>> And most people say: "hg is faster; hg is better;
>> what's difference anyway between hg and bzr?"
> 
> It would be good to get them playing with packs.
> 
> hg *was* faster at everything; we're now faster (for me) for a number of
> things such as commit, and as folk keep pushing our infrastructure
> improvements out to each command this will improve further.
> 
> Are you asking about the difference, or are the people commenting
> asking?

Here the numbers on win32 with timeit.exe program (suggested by Andrew 
Voznytsa). Test was done by Alexander Solovyov.

Testing batch file:

@echo off

cd bzrcx
timeit C:\Python25\python.exe C:\Python25\Scripts\bzr init 
--knitpack-experimental .
timeit C:\Python25\python.exe C:\Python25\Scripts\bzr add -q
timeit C:\Python25\python.exe C:\Python25\Scripts\bzr ci -q -m "initial"
cd ..
timeit C:\Python25\python.exe C:\Python25\Scripts\bzr clone -q bzrcx bzr1
timeit C:\Python25\python.exe C:\Python25\Scripts\bzr clone -q bzrcx bzr2
timeit C:\Python25\python.exe C:\Python25\Scripts\bzr clone -q bzrcx bzr3

cd hgcx
timeit hg init .
timeit hg add -q
timeit hg ci -q -m "initial"
cd ..
timeit hg clone -q hgcx hg1
timeit hg clone -q hgcx hg2
timeit hg clone -q hgcx hg3


Here the numbers for bzr (time in seconds):

time: 0.891
time: 0.937
time: 1.313

time: 2.344
time: 2.031
time: 2.062

Here the numbers for hg:

time: 0.282
time: 0.422
time: 1.531

time: 1.937
time: 1.453
time: 1.593

Alexander Solovyov said he completely satisfied by this bzr result,
and think this numbers acceptable for him. Also he's interested
in speed over network with our smart server.
He said: "if only bzr had such speed six month ago, I would never
look at Mercurial".

Note: this comparison is not fair enough, because Alexander ran 
standalone hg.exe and python-based bzr. In my testing running
standalone bzr.exe is faster, because start-up time of bzr.exe
is lower. On moderate machine difference is about 100-300 ms.
Though loading plugins costs much bigger -- about 500 ms for
my set of plugins.

Alexander.



More information about the bazaar mailing list