ffc-server-python-sdk 1.1.0

A Python Server SDK for featureflag.co project

ffc-server-python-sdk

Introduction

This is the Python Server SDK for the feature management platform featureflag.co. It is intended for use in a multiple-users python server applications.

This SDK has two main purposes:

• Store the available feature flags and evaluate the feature flags by given user in the server side SDK
• Sends feature flags usage, and custom events for the insights and A/B/n testing.

Data synchonization

We use websocket to make the local data synchronized with the server, and then store them in the memory by default. Whenever there is any changes to a feature flag or his related data, the changes would be pushed to the SDK, the average synchronization time is less than 100 ms. Be aware the websocket connection can be interrupted by any error or internet interruption, but it would be restored automatically right after the problem is gone.

Offline mode support

In the offline mode, SDK DOES not exchange any data with featureflag.co, this mode is only use for internal test for instance.

To open the offline mode:

config = Config(env_secret, offline=True)

Evaluation of a feature flag

SDK will initialize all the related data(feature flags, segments etc.) in the bootstrapping and receive the data updates in real time, as mentioned in the above

After initialization, the SDK has all the feature flags in the memory and all evaluation is done locally and synchronously, the average evaluation time is < 10 ms.

Installation

install the sdk in using pip, this version of the SDK is compatible with Python 3.6 through 3.9.

pip install ffc-server-python-sdk

SDK

Applications SHOULD instantiate a single instance for the lifetime of the application. In the case where an application needs to evaluate feature flags from different environments, you may create multiple clients, but they should still be retained for the lifetime of the application rather than created per request or per thread.

Bootstrapping

The bootstrapping is in fact the call of constructor of FFCClient, in which the SDK will be initialized, using streaming from featureflag.co.

The constructor will return when it successfully connects, or when the timeout(default: 15 seconds) expires, whichever comes first. If it has not succeeded in connecting when the timeout elapses, you will receive the client in an uninitialized state where feature flags will return default values; it will still continue trying to connect in the background unless there has been a network error or you close the client(using stop()). You can detect whether initialization has succeeded by calling initialize().

The best way to use the SDK as a singleton, first make sure you have called ffcclient.set_config() at startup time. Then ffcclient.get() will return the same shared ffcclient.client.FFCClient instance each time. The client will be initialized if it runs first time.

from ffcclient.config import Config
from ffcclient import get, set_config 

set_config(Config(env_secret))
client = get()

if client.initialize:
    # your code

You can also manage your ffcclient.client.FFCClient, the SDK will be initialized if you call ffcclient.client.FFCClient constructor.

from ffcclient.config import Config
from ffcclient.client import FFCClient

client = FFCClient(Config(env_secret), start_wait=15)

if client.initialize:
    # your code

If you prefer to have the constructor return immediately, and then wait for initialization to finish at some other point, you can use ffcclient.client.FFCClient.update_status_provider object, which provides an asynchronous way, as follows:

from ffcclient.config import Config
from ffcclient.client import FFCClient

client = FFCClient(Config(env_secret), start_wait=0)
if client._update_status_provider.wait_for_OKState():
    # your code

Evaluation

SDK calculates the value of a feature flag for a given user, and returns a flag vlaue/an object that describes the way that the value was determined.

User: A dictionary of attributes that can affect flag evaluation, usually corresponding to a user of your application. This object contains built-in properties(key, name, email and country). The only mandatory property is the key, which must uniquely identify each user; this could be a username or email address for authenticated users, or a ID for anonymous users.

All other built-in properties are optional, it's strongly recommended to set name in order to search your user quickly You may also define custom properties with arbitrary names and values.

if client.initialize:
    user = {'key': user_key, 'name': user_name, 'age': age}
    flag_value = client.variation(flag_key, user)
    # your if/else code according to flag value

If evaluation called before SDK client initialized or you set the wrong flag key or user for the evaluation, SDK will return the default value you set. The ffcclient.common_types.FlagState will explain the details of the last evaluation including error raison.

If you would like to get variations of all feature flags in a special environment, you can use ffcclient.client.FFCClient.get_all_latest_flag_variations, SDK will return ffcclient.common_types.AllFlagStates, that explain the details of all feature flags

if client.initialize:
    user = {'key': user_key, 'name': user_name}
    all_flag_values = client.get_all_latest_flag_variations(user)
    ed = all_flag_values.get(flag_key)
    flag_value = ed.variation
    # your if/else code according to flag value

    

Experiments (A/B/n Testing)

We support automatic experiments for pageviews and clicks, you just need to set your experiment on our SaaS platform, then you should be able to see the result in near real time after the experiment is started.

In case you need more control over the experiment data sent to our server, we offer a method to send custom event.

client.track_metric(user, event_name, numeric_value);

numeric_value is not mandatory, the default value is 1.

Make sure track_metric is called after the related feature flag is evaluated by simply calling variation or variation_detail otherwise, the custom event won't be included into the experiment result.