Git svn clone: How to defer fetch of revision history

I often have the case that I want to work on a SVN repository right away. But an ordinary git svn clone [url] also clones the entire history. So I want to speed things up. The first part is to fetch only the last revision into your Git repository. I do it like so:

REV=`svn info $URL |grep Revision: | awk '{print $2}'`

git svn clone -r$REV:HEAD $URL $PROJECT_FOLDER

(more info in the StackOverflow article: "How to git-svn clone last n revisions from svn"

This way I'm up and running and can work immediately. But without local copy of the history.

The question is, how do I afterwards fetch history from the svn repository?

And preferably, can this be done in chunks of, say 1000 revisions (in reverse order). Any help here would be greatly appreciated :)


I found out how it can be done. The trick is not to use git svn clone. Instead, use git svn init and git svn fetch individually. Modified the example:

REV=`svn info $URL |grep Revision: | awk '{print $2}'`

cd !$ #goes into dir named $PROJECT_FOLDER
git svn init -s $URL #-s implies --stdlayout with /trunk /tags /branches
git svn fetch -r $REV

# hack, hack, hack

# or update history (fetch 50 revisions back each loop
for (( r=$REV; r>0; r-=50 )); 
  git svn fetch -r $r:HEAD

None of the suggested answers will work. git svn fetch with a revision will only retrieve newer revisions than what is already cloned. You may be able to use git svn reset to go back to an older revision and retrieve from there, but you'll have to do some dirty work afterwards to 'graft' your newer revisions back onto the full tree (the SHA1 of an SVN revision in git depends on the entire parentage of the revision). If you're handy with the scalpels git offers you, go for it.

It's much easier to just avoid the issue.

  • Do an initial clone of the last few revisions, so you can get working immediately;
  • Start another clone of the full history into another directory/git repository;
  • Work in your partial history as much as you want;
  • When the full clone completes, use an approach like to copy your work from the partial repository to the full one.

So, that's a partial answer - how can you afterwards fetch history? Fetch it into another repo and copy what you need over. Can it be done in chunks of 1000 in reverse order? With the scalpels, and a lot of patience, it could, but it's unlikely worth it. The full fetch running forward is going to outrun the overhead of all those first revisions grabbed by each block you git svn fetch, and the fixup will get tedious.

git svn fetch appears to "remember" the revisions it's seen previously. I've been having success with doing ranges:

git svn fetch -r 0:100
git svn fetch -r 100:200
git svn fetch -r 4500
git svn rebase
git svn fetch -r 200:300

I fetched the most recent revisions and then started "filling in" the gaps. It seems to work fine.

Jesper -- if your repository doesn't have a revision 1000, then there's nothing for it to fetch. Make sure the revision numbers you're using are valid!


 ? jQuery draggable clone with y axis restriction
 ? Universal object cloning solution
 ? Clone (deep copy) Entity LINQ
 ? How can I find copy/paste (duplicate, clone) code in Perl?
 ? C++ - cloning base class
 ? Which Ruby classes support .clone?
 ? Duplicate a variable in AS3
 ? how to make the generice data type accessable to clone() method
 ? Is clone() method is equivalent to pass-by-value concept in Java
 ? java Cloneable: Base class implements it incorrectly; what to do?