Python Air Flags SDK

Feature Toggles (often also refered to as Feature Flags) are a powerful technique, allowing teams to modify system behavior without changing code. Pete Hodgson

Air flag definitions

We allow 2 different type of configuration files for the Air flags definitions:

YAML

flag_name:
    value: true
    description: This an amazing Air flag with scheduled rollout
    rollout:
        strategy: scheduled
        percentage: 50
        start_date: 2021-11-01
        end_date: 2021-12-01
other_flag:
    value: true
    description: This is an actived flag
sel_flag:
    value: true
    description: This is an actived flag
    selectived: ["id-1234", "id-5678"]

JSON

{
    "flag_name": {
        "value": true,
        "description": "This is an amazing Air flag with progressive rollout",
        "rollout": {
            "strategy": "progressive",
            "percentage": 5,
            "start_date": "2021-11-01",
            "end_date": "2021-12-01"
        }
    },
    "other_flag": {
        "value": true,
        "description": "This is other flag with a static rollout percentage",
        "rollout": {
            "strategy": "canary",
            "percentage": 50
        }
    },
    "other_one": {
        "value": false,
        "description": "This is an inactived flag"
    }
}

Flag:

Field	Context	Type	Required
flag_name	The flag_name is the name of the flag	str	Yes
value	Status of the flag	bool	Yes
description	Short description of the flag	str	No
selectived	Identifiers selectives for the flag	str / list	No
rollout	Custom object to set up	object	No

Canary rollout:

Field	Context	Type	Required
strategy	Rollout strategy 'canary'	str	Yes
percentage	Percent of traffic to roll the feature out [0 : 100] 'only applies to active flags'	int	Yes
start_date	Start date of the flag with format 'YYYY-mm-dd'	date / str	No
end_date	Expiration date of the flag with format 'YYYY-mm-dd'	date / str	No

Scheduled rollout:

Field	Context	Type	Required
strategy	Rollout strategy 'scheduled'	str	Yes
percentage	Percent of traffic to roll the feature out [0 : 100] 'only applies to active flags'	int	No
start_date	Start date of the flag with format 'YYYY-mm-dd'	date / str	Yes
end_date	Expiration date of the flag with format 'YYYY-mm-dd'	date / str	Yes

Progressive rollout:

Field	Context	Type	Required
strategy	Rollout strategy 'progressive'	str	Yes
percentage	Percent of traffic to start the rollout, it will increase progressively until 100 at end_date	int	Yes
start_date	Start date of the flag with format 'YYYY-mm-dd'	date / str	Yes
end_date	Expiration date of the flag with format 'YYYY-mm-dd'	date / str	Yes

Usage

import air_flags

flags = air_flags.init("/path/to/af_definition.json")

# Used as boolean statement
if flags.flag_name:
    print("Oh yeah!")

# Used as boolean statement with selectived
client_id = "id-1234"
if flags.flag_name(client_id):
    print("Oh yeah!")

# Used as decorator
@flags.is_active("flag_name")
def order_beer():
    pass

# Used as decorator with selectived
client_id = "id-1234"
@flags.is_active("flag_name", client_id)
def order_beer():
    pass

Best practices and considerations

Increased technical debt is a common objection to implementing feature flags.

• Standardize naming:
  • Naming your flags to be as much descriptive as possible.
  • Use prefixes to categorize / group them.
  • Include a description with expected the behavior.
• Priorities using selectived checks:
  • Active if the identifier is selectived.
  • Check rollout and value if the identifier is not selectived.
• Using rollout can generate different value outputs, be careful re-using the same flag in different parts of your application.

Contributing to the SDK

Please refer to CONTRIBUTING.md.

License

Licensed under the BSD license, see LICENSE