On Thu, May 01, 2003 at 12:36:17PM -0500, cmpilato@collab.net wrote:
> mark benedetto king <mbk@boredom.org> writes:
>
> > On Thu, May 01, 2003 at 11:33:20AM -0500, cmpilato@collab.net wrote:
> > > Karl Fogel <kfogel@newton.ch.collab.net> writes:
> > >
> > > > cmpilato@collab.net writes:
> > > > > This I disagree with. I had always just assumed that we would be
> > > > > adding nothing but a single svn_wc_entry_t member (repos_url), and a
> > > > > single XML attribute to match that (repos-url). Determining the
> > > > > repos_path from the repos_url and url is trivial string math.
> > > >
> > > > Does it really matter? Both ways work, each is optimized for
> > > > different circumstances...
> > >
> > > It matters in terms of code churn. You could add support for storing
> > > this repos_url member today without effecting the rest of the WC code
> > > at all.
> >
> > It makes for twice as much entries-rewriting for switch --relocate, not
> > that it really matters, since the entries-file will only be written once.
>
> Huh? Entries files are read and written in full. It doesn't cost
> non-trivially more to:
>
> read and parse a whole file, change two pieces of data, then unparse
> and write the whole file
>
> than it does to:
>
> read and parse a whole file, change one piece of data, then unparse
> and write the whole file
>
> That said, I'm not the guy writing the code, so dude can do whatever
> he sees fit to do.
That's exactly what I said. :-)
I was pointing out that there will be (trivial) churn in the code in the
issues-951-dev branch.
--ben
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu May 1 19:46:46 2003