Bazaar Fast Import
bzr-fastimport provides a back-end for fast data importing into Bazaar. The usual recipe is:
bzr init-repo . front-end | bzr fast-import -
Key features include:
- Fast data importing from most popular VCS tools
- Easy to develop custom exporters
- Import whole repositories: branches are created for each head detected.
Note: bzr-fastimport was inspired by Shawn Pearce's git-fast-import and uses a stream format that is upwardly compatible with it. Thanks go to Shawn and the Git community for developing that tool and a range of matching front-ends.
Downloads, Bugs, etc.
This plugin is hosted on Launchpad here: https://launchpad.net/bzr-fastimport. After installing Bazaar, grab the latest source like this:
bzr branch lp:bzr-fastimport fastimport
Bugs can be raised and tracked via Launchpad's bug tracker: https://bugs.launchpad.net/bzr-fastimport/. If you have questions or ideas for discussion, please use the main Bazaar mailing list, <email@example.com>.
After installing the plugin, see the online help and the bundled documentation. A summary of the main features is given below.
The numbers are very dependent on the number of files and number of revisions. As a rough guide, around 1000 revisions per minute can be imported for projects of several hundred files on a typical desktop. For example, the 6000+ revisions of Wordpress (500 files) can be imported from a Subversion export in just over 2 minutes on my Dell Dimension 9200. Other benchmarking suggests import performance 5-20 times that of bzr-git, for example.
See Bazaar Fast Import Front-Ends for a list of recommended exporters from other tools. Front-ends are available for Subversion, CVS, Git, Mercurial, Darcs and Perforce among others.
Many projects take the opportunity to clean up cruft when migrating from an older VCS tool to a new one. Rolling your own exporter is often a good way of doing that. If you want to do this, see Bazaar Fast Import Stream Format for the detailed format specification. Exporters can easily be developed in most scripting languages (Python, Perl, Ruby, etc.) and are typically just a few hundred lines long.
If an import is interrupted or unexpectedly crashes, it can be restarted. Technically, inside the destination repository's control directory a file named fastimport-id-map is kept that holds the mapping from import stream commit-ids to Bazaar revision-ids. That file is saved at each checkpoint and on completion. When fast-import runs, the repository is checked for consistency against that file (if present) and revisions already loaded are skipped in the import stream accordingly.
Mirroring of Foreign Repositories
As long as the fastimport-id-map file is not deleted and the front-end generates an export which begins with the same content as the previous import, bzr-fastimport can be used to maintain a Bazaar mirror of a foreign repository.
Whole Repository Imports
If run inside a shared repository, branches will be created or updated as required.
If run inside a standalone tree, all revisions from the repository will be imported, the current branch will be made the trunk and information is output about the lost heads.