dvc

1. Bug fixes;
2. Added dvc install for installing git post-checkout hook;
3. Support dvc checkout for a single file;
4. Support for reflink/hardlink/symlink/copy types of cache links;
5. Support for configurable and shareable cache dirs;
6. Stages without dependencies are now ran by dvc repro;
7. Dvc now checks for updates once a week and prints a warning if newer version is available;

1. Config schema verification
2. Clouds in config are replaced with remotes. Introduced special command dvc remote for adding/modifying/deleting remotes. dvc push/pull/fetch/status now accepts -r|--remote argument for specifying the particular remote repository you want to use(use dvc config core.remote myremote to setup default remote repo).
3. GCP cloud no longer requires ProjectName, as long as your credentials are setup correctly and gsutil works for you.
4. DVC version is now generated dynamically, depending on whether it is release version or from the git(e.g. dvc 0.9.5+89ed2c.mod).
5. Fixed automatic dvcfile name generation for dvc run -o directory/.
6. Both dvcfiles and configs are no longer written empty parameters to.
7. Test coverage increased to 91%.
8. Experimental SSH cloud support added.
9. Cloud drivers now use local state to obtain md5, instead of recomputing it every time.
10. Introduced local config(.dvc/config.local).
11. AWS cloud migrated from boto to boto3.
12. Introduces 'md5' field for dvcfiles, which allows us to detect if dvcfile itself has changed(i.e. cmd).
13. Non-binary dependencies are now converted to POSIX format(CRLF -> CR) while computing md5, so that dependencies which point to files tracked by git are now compatible on windows and *nix.
14. Introduced dvc metrics for reading metrics on all branches.

Use POSIX path on windows and convert to POSIX for backward compatibility.

1. Fixed bug in data cloud key creation on Windows.
2. Added dynamic versions(using commit SHA) for dvc built from upstream repo.
3. Added long package description for pypi.

Pypi package fixes.

DVC beta release.

Make your data science projects reproducible and shareable.

Includes:

• Import files, http links and drupbox links
• Command reproduction: recursion, force-reproduction
• Data synchronization: AWS S3 and GCP Storage
• Data item locks
• End-to-end machine learning tutorial: stackoverflow tag classification