RE: Re: [viewvc-dev] user alternative when row_limit is exceeded
David Carson <[email protected]> Wed, 4 May 2011 11:11:27 -0700 (PDT)
| Newsgroups | gmane.comp.version-control.cvs.viewcvs.devel |
|---|---|
| Message-ID | <1621170253.18609.1304532687649.JavaMail.httpd__46058.326905086$1304532696$gmane$org@localhost> |
> On 05/03/2011 05:46 PM, David Carson wrote: > > I started looking at the SQL query itself. This reveals to me the fact > > that because this tool was designed back in the days of CVS, the DB is > > organized by commits as individual file changes. In other words, one SVN > > commit with 3,000 files added creates 3,000 commit rows instead of one > > new row. > > David, this observation is, in my opinion, the single most important one you > could have made. Subversion support in the commits database is a bolt-on -- > a contributor's quick work to make Subversion commits searchable while still > maintaining DB schema compatibility with the old CVS-centric Bonsai product > -- and that results in a number of interesting side effects: > > - no commit history for directories > - excess metadata storage > - insufficient handling of copies > - zero support for branches/tags > - etc. > > There is zero version-control-system abstraction in the cvsdb code such as > you might find in the general "vclib" submodule. > > Is the solution here to design a new schema that's suitable for Subversion > and can solve as many of these shortcomings as possible? > I would certainly agree that this is the correct long-term solution. Are you suggesting there be multiple schemas or just a new schema replacing the old one, which would better suit SVN but could also handle CVS? ------------------------------------------------------ http://viewvc.tigris.org/ds/viewMessage.do?dsForumId=4251&dsMessageId=2730178 To unsubscribe from this discussion, e-mail: [[email protected]].