Skip to main content

Python SDK reference

note

This SDK is not currently supported on Harness Self-Managed Enterprise Edition (on premises).

This topic describes how to use the Harness Feature Flags Python SDK for your Python application.

For getting started quickly, you can use our sample code from the Python SDK README. You can also clone and run a sample application from the Python SDK GitHub Repository.

Before You Begin

You should read and understand the following:

Version

Latest SDK version can be found on GitHub Release Page

Requirements

To use this SDK, make sure you:  

Install the SDK

Install the python SDK using pip

python -m pip install harness-featureflags

Initialize the SDK

To initialize the Python SDK, you need to:

  1. Add your Server SDK Key to connect to your Harness Environment.
  2. Add a Target that you want to Evaluate against a Feature Flag.
  3. (Optional) Configure the SDK options. For more details on what features you can configure for this SDK, go to Configure the SDK.
note

If the SDK fails to authenticate with the Feature Flags service and cannot initialise, the SDK will operate in a "limited" mode and will only serve the default values you provide in variation calls.

Block initialization

By default, when initializing the Harness Feature Flags client, the initialization process is non-blocking. This means that the client creation call returns immediately, allowing your application to continue its startup process without waiting for the client to be fully initialized. If you evaluate a flag before the client has finished initializing, the default variation you provided can be returned as the evaluation result, because the SDK has not finished caching your remote Flag configuration stored in Harness.

You can choose to wait for the client to finish initializing before continuing. To achieve this, you can call the wait_for_initialzation method, which blocks until the client is fully initialized. Example usage:

client = CfClient(api_key)
# Block the thread until all flag configuration has been loaded into the cache.
client.wait_for_initialization()
result = client.bool_variation('identifier_of_your_bool_flag', target, False)

Add the Server SDK Key

To connect to the correct Environment that you set up on the Harness Platform, you need to add the Server SDK Key from that Environment. Input the Server SDK Key into the api_key parameter. For example:

"""  
Put the API Key here from your environment
"""
api_key = "YOUR_API_KEY";

cf = CfClient(api_key);

Add a Target

Details

What is a Target? Targets are used to control which users see which Variation of a Feature Flag, for example, if you want to do internal testing, you can enable the Flag for some users and not others. When creating a Target, you give it a name and a unique identifier. Often Targets are users but you can create a Target from anything that can be uniquely identified, such as an app or a machine.

For more information about Targets, go to Targeting Users With Flags.

To add a Target, build it and pass in arguments for the following:

ParameterDescriptionRequired?Example
identifierUnique ID for the Target.Read Regex requirements for Target names and identifiers below for accepted characters.Requiredidentifier='HT_1'
nameName for this Target. This does not have to be unique. Note: If you don’t provide a value, the name will be the same as the identifier.Read Regex requirements for Target names and identifiers below for accepted characters.OptionalNote: If you don't want to send a name, don't send the parameter. Sending an empty argument will cause an error.name="Harness_Target_1"
attributesAdditional data you can store for a Target, such as email addresses or location.Optionalattributes={"email": "demo@harness.io"}
Regex requirements for Target names and identifiers

Identifier

Regex: ^[A-Za-z0-9.@_-]*$
Must consist of only alphabetical characters, numbers, and the following symbols:
. (period)
@ (at sign)
-(dash)
_ (underscore)

The characters can be lowercase or uppercase but cannot include accented letters, for example Cafe_789.

Name Regex: ^[\\p{L}\\d .@_-]*$

Must consist of only alphabetical characters, numbers, and the following symbols:
. (period)
@ (at sign)
-(dash)
_ (underscore)
(space)

The characters can be lowercase or uppercase and can include accented letters, for example Café_123.

For example:

 target = Target(identifier='HT_1', name="Harness_Target_1", attributes={"email": "demo@harness.io"})

Configure the SDK

You can configure the following features of the SDK:

NameExampleDescriptionDefault Value
baseUrlwith_base_url("https://config.ff.harness.io/api/1.0")The URL used to fetch Feature Flag Evaluations. When using the Relay Proxy, change this to: http://localhost:7000https://config.ff.harness.io/api/1.0
eventUrlwith_events_url("https://events.ff.harness.io/api/1.0")The URL for posting metrics data to the Feature Flag service. When using the Relay Proxy, change this to: http://localhost:7000https://events.ff.harness.io/api/1.0
pollIntervalConfig(pull_interval=60)The interval in seconds that we poll for changes when you are using stream mode.60 (seconds)
streamEnabledwith_stream_enabled(True)Set to True to enable streaming mode. Set to False to disable streaming mode.True
analyticsEnabledwith_analytics_enabled(True)Set to True to enable analytics. Set to False to disable analytics.
Note: When enabled, analytics data is posted every 60 seconds.
True
maxAuthRetrieswith_max_auth_retries(10)The number of retry attempts to make if client authentication fails on a retryable HTTP error10

For example:

# Create a Feature Flag Client  
client = CfClient(apiKey,
with_base_url("https://config.ff.harness.io/api/1.0"),
with_events_url("https://events.ff.harness.io/api/1.0"),
with_stream_enabled(True),
with_analytics_enabled(True),
Config(pull_interval=60))

Evaluate a Flag

Evaluating a Flag is when the SDK processes all Flag rules and returns the correct Variation of that Flag for the Target you provide.

If a matching Flag can’t be found, or the SDK can’t remotely fetch flags, the default value is returned, and an error is logged.

There are different methods for the different Variation types and for each method you need to pass in:

  • Identifier of the Flag you want to evaluate
  • The Target object you want to evaluate against
  • The default Variation

For example:

Evaluate a boolean Variation

result = cf.bool_variation("identifier_of_your_bool_flag", target, False);  

Evaluate a string Variation

result = client.string_variation('identifier_of_your_string_flag', target, "")

Evaluate a number Variation

We recommend using the newer int_or_float_variation method, which replaces int_variation and number_variation .

result = client.int_or_float_variation('identifier_of_your_number_flag', target, -1)

Evaluate a JSON Variation

client.json_variation('identifier_of_your_json_flag', target, {})
note

If you evaluate a feature flag when initialization fails, the default variation you provided is returned as the evaluation result.

Check the type of a flag

If you would like to know the type of a flag at any point, you can use the method get_flag_type

This is helpful if you don't know the types of flags up front, or if you would like to create a type agnostic wrapper.

Note, this feature is specific to the Python SDK at present, and we have no current plans of rolling it out to other SDK languages.

# Get the flag type
flag_type = client.get_flag_type(flag_identifier)
log.info("Flag '%s' is of type '%s'", flag_identifier, flag_type)

Flag types are defined as followed:

class FeatureFlagType(str, Enum):
BOOLEAN = "boolean"
INT_OR_FLOAT = "int"
STRING = "string"
JSON = "json"
FLAG_NOT_FOUND = "flag_not_found"

For a full example, see the Python SDK GitHub repository

Test your app is connected to Harness

When you receive a response showing the current status of your Feature Flag, go to the Harness Platform and toggle the Flag on and off. Then, check your app to verify if the Flag Variation displayed is updated with the Variation you toggled.

note

The SDK must run for at least 60 seconds before it sends metrics. Please ensure metrics have not been disabled in the SDK.

Close the SDK Client

In most applications, you won't need to close the SDK client.

However, you should close the SDK client if:

  • Your application is about to terminate. Closing the client ensures that all associated resources are released.
  • You have determined that you do not need to evaluate flags again in your application lifecycle.
important

The SDK does not evaluate flags after the client is closed.

To close the SDK client:

  • Assuming you have initialized an SDK client instance named client, call the following function:

    client.close()

Additional options

Configure your logger

The SDK provides a logger that wraps the standard Python logging package. You can import and use it with the following:

from featureflags.util import log  
log.info("Hello, World!")

To change the default log level, you can use the standard logging levels

from featureflags.util import log  
import logging

log.setLevel(logging.WARN)

Use the Relay Proxy

When using your Feature Flag SDKs with a Harness Relay Proxy you need to change the default URL and events URL to http://localhost:7000 when initializing the SDK. To do this:

  1. Import the URL helper functions, for example:
from featureflags.config import with_base_url  
from featureflags.config import with_events_url
  1. Pass the new URLs in when initializing the SDK, for example:
    client = CfClient(api_key,  
with_base_url("https://config.feature-flags.uat.harness.io/api/1.0"),
with_events_url("https://event.feature-flags.uat.harness.io/api/1.0"))

Sample code for a Python application

Here is a sample code for integrating with the Python SDK:

import time  

from featureflags.evaluations.auth_target import Target
from featureflags.client import CfClient
from featureflags.util import log
from featureflags.config import with_base_url
from featureflags.config import with_events_url


def main():
log.debug("Starting example")
api_key = "Your API key"
client = CfClient(api_key,
with_base_url("https://config.ff.harness.io/api/1.0"),
with_events_url("https://events.ff.harness.io/api/1.0"))

target = Target(identifier='HT_1', name="Harness_Target_1", attributes={"location": "emea"})

while True:
result = client.bool_variation('harnessappdemodarkmode', target, False)
log.debug("Result %s", result)
time.sleep(10)

if __name__ == "__main__":
main()

Troubleshooting

The SDK logs the following codes for certain lifecycle events, for example authentication, which can aid troubleshooting.

CodeDescription
1000Successfully initialized
1001Failed to initialize due to authentication error
1002Failed to initialize due to a missing or empty API key
1003wait_for_initialzation was called and the SDK is waiting for initialzation to complete
2000Successfully authenticated
2001Authentication failed with a non recoverable error
2002Authentication failed and is retrying
2003Authentication failed and max retries have been exceeded
3000SDK Closing
3001SDK closed successfully
4000Polling service started
4001Polling service stopped
4002Poller has fetched flags and groups from backend successfully
5000Streaming service started
5001Streaming service stopped
5002Streaming event received
5003Streaming disconnected and is retrying to connect
5004Streaming stopped
5005Stream is still retrying to connect after 4 attempts
6000Evaluation was successfully
6001Evaluation failed and the default value was returned
7000Metrics service has started
7001Metrics service has stopped
7002Metrics posting failed
7003Metrics posting success
7004Metrics max target size exceeded
7005Metrics batch targets sending success
7006Metrics batch targets sending failed
7007Metrics max evaluation size reached
8005Fetching flag by identifier request failed and is retrying
8006Fetching group by identifier request failed and is retrying
8007Fetching all flags request failed and is retrying
8008Fetching all groups request failed and is retrying
8009Fetching flag by identifier request failed
8010Fetching group by identifier request failed
8011Fetching all flags request failed
8012Fetching all groups request failed