datafs 0.6.2

DataFS is an abstraction layer for data storage systems. It manages file versions and metadata using a json-like storage system like AWS's DynamoDB and relies on PyFilesystem to abstract file storage, allowing you to store files locally and on the cloud in a seamless interface.

DataFS Data Management System

DataFS is a package manager for data. It manages file versions, dependencies, and metadata for individual use or large organizations.

Configure and connect to a metadata Manager and multiple data Services using a specification file and you’ll be sharing, tracking, and using your data in seconds.

• Free software: MIT license

• Documentation: https://datafs.readthedocs.io.

Features

• Explicit version and metadata management for teams

• Unified read/write interface across file systems

• Easily create out-of-the-box configuration files for users

• Track data dependencies and usage logs

• Use datafs from python or from the command line

• Permissions handled by managers & services, giving you control over user access

Usage

First, configure an API. Don’t worry. It’s not too bad.

We’ll assume we already have an API object created and attached to a service called “s3”. Once you have this, you can start using DataFS to create and use archives.

$ datafs create my_new_data_archive --description "a test archive"
created versioned archive <DataArchive s3://my_new_data_archive>

$ echo "initial file contents" > my_file.txt

$ datafs update my_new_data_archive my_file.txt

$ datafs cat my_new_data_archive
initial file contents

Versions are tracked explicitly. Bump versions on write, and read old versions if desired.

$ echo "updated contents" > my_file.txt

$ datafs update my_new_data_archive my_file.txt --bumpversion minor
uploaded data to <DataArchive s3://my_new_data_archive>. version bumped 0.0.1 --> 0.1.

$ datafs cat my_new_data_archive
updated contents

$ datafs cat my_new_data_archive --version 0.0.1
initial file contents

Pin versions using a requirements file to set the default version

$ echo "my_new_data_archive==0.0.1" > requirements_data.txt

$ datafs cat my_new_data_archive
initial file contents

All of these features are available from (and faster in) python:

>>> import datafs
>>> api = datafs.get_api()
>>> archive = api.get_archive('my_new_data_archive')
>>> with archive.open('r', version='latest') as f:
...     print(f.read())
...
updated contents

If you have permission to delete archives, it’s easy to do. See administrative tools for tips on setting permissions.

$ datafs delete my_new_data_archive
deleted archive <DataArchive s3://my_new_data_archive>

See examples for more extensive use cases.

Installation

pip install datafs

Additionally, you’ll need to choose a manager and services:

Managers:

• MongoDB: pip install pymongo

• DynamoDB: pip install boto3

Services:

• Ready out-of-the-box:

  • local

  • shared

  • mounted

  • zip

  • ftp

  • http/https

  • in-memory

• Requiring additional packages:

  • AWS/S3: pip install boto

  • SFTP: pip install paramiko

  • XMLRPC: pip install xmlrpclib

Requirements

For now, DataFS requires python 2.7. We’re working on 3x support.

Todo

See issues to see and add to our todos.

Credits

This package was created by Justin Simcock and Michael Delgado of the Climate Impact Lab. Check us out on github.

Major kudos to the folks at PyFilesystem. Thanks also to audreyr for the wonderful cookiecutter package, and to Pyup, a constant source of inspiration and our silent third contributor.

History

0.1.0 (2016-11-18)

• First release on PyPI.