bev 0.5.0

A small manager for versioned data

Flexible version control for files and folders.

Install

The simplest way is to get it from PyPi:

pip install bev

Or if you want to try the latest version from GitHub:

git clone https://github.com/neuro-ml/bev.git
cd bev
pip install -e .

# or let pip handle the cloning:
pip install git+https://github.com/neuro-ml/bev.git

Getting started

1. Choose a folder for your repository and create a basic config (.bev.yml):

main:
  storage: /path/to/storage/folder

meta:
  hash: sha256

2. Run init

bev init

3. Add files to bev

bev add /path/to/some/file.json .
bev add /path/to/some/folder/ .
bev add /path/to/some/image.png .

4. ... and to git

git add file.json.hash folder.hash image.png.hash
git commit -m "added files"

5. Access the files from python

import imageio
from bev import Repository

# `version` can be a commit hash or a git tag 
repo = Repository('/path/to/repo', version='8a7fe6')
image = imageio.imread(repo.resolve('image.png'))

Advanced usage

Here are some tutorials that cover more advanced configuration, including multiple storage locations and machines:

1. Create a repository - needed only at first time setup
2. Adding files
3. Accessing files

Why not DVC?

DVC is a great project, and we took inspiration from it while designing bev. However, out lab has several requirements that DVC doesn't meet:

1. Our data caches are spread across multiple HDDs - we need support for multiple cache locations
2. We have multiple machines, and each of them has a different storage configuration: locations, number of HDDs, their volumes - we need a flexible way of choosing the right config depending on the machine
3. Often we simultaneously conduct experiments on different versions of the same data - we need easy access to multiple version of the same data
4. The need for dvc checkout after git checkout is error-prone, because it can lead to situations when the data is not consistent with the current commit - we need a more constrained relation between data and git

bev supports all four out of the box!

However, if these requirements are not essential to your project, you may want to stick with DVC - its community and tests coverage is much larger.