pombredanne / pyrollout

Python feature flagging

pyrollout

Python feature flagging

Inspired by the rollout Ruby gem by James Golick and a Python port of it called proclaim by [Curt Micol](Curt Micol) which I found after reading Using Feature Flags to Ship Changes with Confidence by Mathias Meyer.

Both of these use redis, and I don't like that requirement. I seek to support multiple storage/persistence mechanisms.

Usage

To initialize pyrollout with the default in-memory feature & user storage:

import pyrollout
rollout = pyrollout.Rollout()

Now add features:

from pyrollout.feature import Feature
# Open to all by using the special group 'ALL'
rollout.add_feature(Feature('feature_for_all', groups=['ALL']))
# Open to select groups
rollout.add_feature(Feature('feature_for_groups', groups=['vip', 'early_adopter]))
# Open to specific user(s), by user ID
rollout.add_feature(Feature('feature_for_users', users=[123, 456, 789]))
# Open to 20% of users, calculated via user ID
rollout.add_feature(Feature('20pct', percentage=20))

Check access to features:

def untested_feature(user):
    # Because this feature was not defined, access will always be denied (by default)
    if not rollout.can(user, 'use_untested_feature'):
        return None
    else:
        do_cool_things()

class FooHandler(BaseHandler):
    def get(self):
        # self.user is a user object in a format pyrollout understands
        if not rollout.can(self.user, 'feature_for_users'):
            self.abort(403)
        else:
            do_foo()

Some feature ideas

• Flag on user property
• Flag on lambda passed the user object
• Use file
• Use memcache, e.g. in Google App Engine
• Use some other persistent database