How exactly does subversion store files in the repository?

I read the subversion book and it is clear to me that subversion does not store individual files but only deltas in order to minimize disk space. Subversion also does the same with binary files as well (this used to be a huge weakness of CVS).

However I do not understand the exact mechanism. When I commit a file what happens?

1. Subversion stores only the diff (and already has the old version)
2. Subversion deletes the previous version, stores the new file intact and creates a reverse diff in order to "re-create" the old version if needed.
3. Something else that I haven't thought of.

The first case might seem the most logical. This however raises another question. If I have in a subversion repository a file with 1000 commits and a new developer checks out a clean copy, then subversion would have to fetch the original version (initial import) and apply 1000 diffs on this before returning the result. Is this correct? Is there some开发者_如何学运维 sort of caching for files where the latest version is kept as well?

Basically where can I find information on the svn repository internals?

Update: Apparently the backend of subversion plays a big role in this. At the time or writing FSFS uses option 1 while BDB uses option 2. Thanks msemack!

Because Subversion's repository format is entirely internal, they are free to change the representation from one revision to the next. I believe the current revision generally stores reverse deltas (your option 2), but also stores complete snapshots periodically so it doesn't have to resolve 1000 diffs before returning a result.

The Subversion 1.6 release notes has a section on Filesystem storage improvements that has some notes on this, and links to other sources. Suffice to say that the details of Subversion data storage are complex and subject to change.

There is also a design document in the Subversion source tree that describes the use of skip deltas in Subversion. Generally, the /notes/ directory contains several useful documents regarding Subversion internals.

From the Subversion Design document (which is quite dated, though) you can get this:

Like many other revision control systems, Subversion stores changes as differences. It doesn't make complete copies of nodes; instead, it stores the latest revision as a full text, and previous revisions as a succession of reverse diffs (the word "diff" is used loosely here – for files, it means vdeltas, for directories, it means a format that expresses changes to directories).

I don't think that was changed since.

Also, see Bubble-Up Method.

I believe the following link would be of assistance to understand the fsfs architecture

http://svn.apache.org/repos/asf/subversion/trunk/subversion/libsvn_fs_fs/structure

the regular FSFS specification might help you.

Or if you use Berkeley DB, here's the specification for that.

FSFS uses reverse deltas to store the changes and skip-deltas to speed up some actions, if I understood everything correctly.

Each time you commit a change, the repository stores a new revision of that overall repository tree, and labels the new tree with a new revision number. Of course, most of the tree is the same as the revision before, except for the parts you changed.

The new revision number is a sequential label that applies to the entire new tree, not just to the files and directories you touched in that revision. However, colloquially, a revision number is used to refer to the change committed in that revision; for example, "the change in r588" ("r588" is shorthand for "revision 588") really means "the difference between repository trees 587 and 588", or put another way, "the change made to tree 587 to produce tree 588".

Have a look at : Subversion FAQ