How to Evaluate Feature Flags in Python

A Python function that evaluates boolean feature flags with user-specific overrides, returning whether a flag is enabled and the reason for the decision.

Easy Python 3.9+ Aug 9, 2026 A/B testing & experimentation 14 views 0 copies

feature flags ab testing experimentation configuration

Python code

54 lines

Python 3.9+

import json

def evaluate_feature_flag(feature_name, context, flag_configs):
    """
    Evaluates a boolean feature flag given a context dictionary.

    Args:
        feature_name: The name of the feature flag.
        context: A dictionary of user/request context (e.g., {"user_id": "123"}).
        flag_configs: A dictionary of flag configurations.

    Returns:
        A tuple (enabled, reason) where enabled is bool and reason explains the decision.
    """
    # Flag not found → disabled with a clear reason
    if feature_name not in flag_configs:
        return False, f"flag '{feature_name}' not found"

    config = flag_configs[feature_name]
    enabled = config.get("enabled", False)

    # Check for user-based override rules
    for rule in config.get("overrides", []):
        if rule.get("user_id") == context.get("user_id"):
            return rule["enabled"], f"override matched user_id={context['user_id']}"

    return enabled, "standard evaluation"


if __name__ == "__main__":
    flags = {
        "dark_mode": {"enabled": True, "overrides": [
            {"user_id": "beta_1", "enabled": False}
        ]},
        "new_checkout": {"enabled": False}
    }

    # Simulate mock evaluation for a normal user
    context_normal = {"user_id": "user_123"}
    result_normal = evaluate_feature_flag("dark_mode", context_normal, flags)
    print(f"dark_mode (normal): {result_normal}")

    # Simulate mock evaluation for an overridden user
    context_beta = {"user_id": "beta_1"}
    result_beta = evaluate_feature_flag("dark_mode", context_beta, flags)
    print(f"dark_mode (beta): {result_beta}")

    # Non-existent flag
    result_missing = evaluate_feature_flag("unknown_flag", context_normal, flags)
    print(f"unknown_flag: {result_missing}")

    # Disabled flag
    result_disabled = evaluate_feature_flag("new_checkout", context_normal, flags)
    print(f"new_checkout: {result_disabled}")

Output

stdout

dark_mode (normal): (True, 'standard evaluation')
dark_mode (beta): (False, 'override matched user_id=beta_1')
unknown_flag: (False, "flag 'unknown_flag' not found")
new_checkout: (False, 'standard evaluation')

How it works

The evaluate_feature_flag function first checks if the flag exists in the config dictionary, returning a disabled result with a clear reason if not found. It then reads the enabled value with .get() so missing keys default to False instead of raising an error. User-specific overrides are checked in order, returning the override result immediately when a matching user_id is found — this makes it easy to test beta users or roll back a feature for specific accounts. The function always returns a tuple of (bool, str), which keeps the decision logic transportable and easy to unit test.

Common mistakes

Forgetting to handle the case where the flag isn't in the config, causing a KeyError
Using `config["enabled"]` directly instead of `.get("enabled", False)` for safe defaults
Checking overrides after evaluating the default enabled state, which can accidentally apply overrides to flags that should always be off

Variations

Use percentage-based rollout with `hash(user_id) % 100 < rollout_percentage`
Store configs in a database and cache them in memory for faster lookups

Real-world use cases

Rolling out a new UI feature gradually to specific user segments while keeping control over edge-case accounts.
A/B testing different checkout flows where a test group gets a new experience and control group stays on the old one.
Emergency kill-switch automation — disable a buggy flag for all users instantly without redeploying code.

How to Evaluate Feature Flags in Python

Python code

Output

How it works

Common mistakes

Variations

Real-world use cases

More from A/B testing & experimentation

Tutorials

Quizzes

Python code

Output

How it works

Common mistakes

Variations

Real-world use cases

More from A/B testing & experimentation

Keep learning

Tutorials

Quizzes