This was originally miscategorized as Feature #2799.
This is a common use case as it prevents the standard trunk/branch/tag development structure.

A shared repository created with @bzr init-repo@ errors when running @ruby script/runner "Repository.fetch_changesets" -e production I get these messages back:@
@bzr: ERROR: Not a branch: "[...]/project_container/.bzr/branch/".@
@bzr: ERROR: Not a branch: "[...]/project_container/.bzr/branch/".@

If we point at a sub-tree, there's still no support for branches.

There are 2 ways to identify the shared repository:

• bzr info will list 'shared repository: .'
• .bzr/repository exists.

If the path points to a shared repo, some file tree should be built.

Branches can be identified by a similar combination of techniques:

• bzr info will list the root repository as 'shared repository'
• .bzr/branch exists.

Sub branches:

• bzr info will list the root repository as 'shared repository' and the parent branch as 'parent branch'

As an example, let's examine the following repo:
[...]

In summary, if the path is a repo, scan the directory tree for directories containing .bzr/branch and run bzr info on those directories. Use that output to find out which branch is the parent and use that information to build the element tree.

I don't know how to do this in ruby, but hopefully someone can take this analysis and put something together.