Today at work at Fronde I needed to create a dump of a ‘project’ inside a repository.  Our SVN setup is to have one repo per team and then within there have a folder for each client/project.  In this case I was providing the backup to a client so it was vital that I gave them just their code and not the whole repo. So my SVN repo was like this:


To dump the clientA history into a portable, re-creatable format, I used svnadmin dump, like this:

svnadmin dump [path to repo] > repo.dump

Which creates a dump of the entire repository into a file called repo.dump.  This took about 10 mins with 1000 versions and used 100% CPU so it would be best to perform this outside of normal work hours.  You have been warned! I then used svndumpfilter (tutorial) to filter just for the ClientA folder (see folder tree above):

svndumpfilter include clientA < repo.dump > clientA.dump

If you have nested repositories, then it breaks with a syntax error.  To get around this you need to run the dump multiple times using the ‘exclude’ directive until you have what you want:

svndumpfilter **exclude** clientB < repo.dump **>>** clientA.dump
svndumpfilter **exclude** clientC < repo.dump **>>** clientA.dump

This didn’t take very long and at the end I had a full svn repository that could be re-created anywhere.  To prove the point I installed SVN on my mac and created the repository.  I then loaded the dump file into it and it worked beautifully ;)

svnadmin create /Users/Dave/ClientA
svnadmin load /Users/Dave/ClientA < ClientA.bak
mkdir /Users/Dave/ClientA-checkout
svn co file:///Users/Dave/ClientA clientA-checkout/

Now that you have checked it, you can delete the whole repo backup file (mine was massive) Repo.svn-dump in my case. Easy as pie once you know how.  The fact that you can’t specify the filter at dump-time is non-intutative, clunky and frankly a waste of space.  However if you only need to do it once it works and produces a very nice result.  Open Source and Open Standards FTW ;)