Performance issues when setting dates from scripts

[Jason Field <jf...@sonaptic.com>]

Hi all,

I've recently been trying to export a VSS repository to Subversion using
the PySvn bindings. Everything seems to be working well, except for the
massive decrease in performance that occurs if I set the script to
change the date revision property of checkins.  Does Subversion require
that ascending revisions have ascending datestamps? I assumed that the
date property was entirely independant of the revision number but
perhaps I'm wrong! The performance drop off makes the repository
unusable unfortunately, and yet it would be very handy to maintain the
entire VSS structure including dates. Any ideas appreciated!

Thanks,

Jason

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org

Re: Performance issues when setting dates from scripts

[Jason Field <jf...@sonaptic.com>]

Following up from my previous post, I think I've discovered what was
going wrong. It seems that either there is a problem with my server's
setup, or that there is some issue between Subversion/PySvn and Apache
that causes the performance degradation. If I run the script using
svnserve as the host, then everything remains at full speed. Everything
is now exported happily from VSS and into Subversion :)

For anyone who is interested in playing with the script I wrote, I've
uploaded it to my personal web server at http://jasonfield.com/ - follow
the freebies link.

Cheers,

Jason

Andrew Thompson wrote:

> Jason Field wrote:
>
>> I've recently been trying to export a VSS repository to Subversion using
>> the PySvn bindings. 
>
>
> pysvn looks like a Python library for SVN. Are you saying that you're
> using a python script that references pysvn to to export your VSS
> data? (huh?)
>
> Or are you're just using pysvn workbench as a user frontend to SVN?
>
>> Everything seems to be working well, except for the
>> massive decrease in performance that occurs if I set the script to
>> change the date revision property of checkins.  Does Subversion require
>> that ascending revisions have ascending datestamps? 
>
>
> Are you saying that if you make two repositories, one including the
> old dates and one excluding the old dates, the performance of the two
> repositories are different? I would consider that very curious...
>
> Wait... Does your "massive decrease in performance" occur during the
> post-conversion usage of the SVN repository, or during the export of
> data from VSS to SVN. If it's only during the export, I would consider
> that a cost of doing business with Microsoft. It *should not* affect
> everyday usage.
>
> > I assumed that the
>
>> date property was entirely independant of the revision number but
>> perhaps I'm wrong! The performance drop off makes the repository
>> unusable unfortunately, and yet it would be very handy to maintain the
>> entire VSS structure including dates. Any ideas appreciated!
>
>
> See recent threads on date ranges over the last few days for the
> ramifications. (I can't recall if it was on users@ or dev@.)
>

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org

Re: Performance issues when setting dates from scripts

[Andrew Thompson <su...@aktzero.com>]

Jason Field wrote:
> I've recently been trying to export a VSS repository to Subversion using
> the PySvn bindings. 

pysvn looks like a Python library for SVN. Are you saying that you're 
using a python script that references pysvn to to export your VSS data? 
(huh?)

Or are you're just using pysvn workbench as a user frontend to SVN?

> Everything seems to be working well, except for the
> massive decrease in performance that occurs if I set the script to
> change the date revision property of checkins.  Does Subversion require
> that ascending revisions have ascending datestamps? 

Are you saying that if you make two repositories, one including the old 
dates and one excluding the old dates, the performance of the two 
repositories are different? I would consider that very curious...

Wait... Does your "massive decrease in performance" occur during the 
post-conversion usage of the SVN repository, or during the export of 
data from VSS to SVN. If it's only during the export, I would consider 
that a cost of doing business with Microsoft. It *should not* affect 
everyday usage.

 > I assumed that the
> date property was entirely independant of the revision number but
> perhaps I'm wrong! The performance drop off makes the repository
> unusable unfortunately, and yet it would be very handy to maintain the
> entire VSS structure including dates. Any ideas appreciated!

See recent threads on date ranges over the last few days for the 
ramifications. (I can't recall if it was on users@ or dev@.)

-- 
Andrew Thompson
http://aktzero.com/
Interested in a hosted SVN repository? Email me, let's talk...

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org