Module aws_lambda_powertools.utilities.feature_flags.feature_flags

Classes

class FeatureFlags (store: StoreProvider, logger: logging.Logger | Logger | None = None)

Evaluates whether feature flags should be enabled based on a given context.

It uses the provided store to fetch feature flag rules before evaluating them.

Examples

from aws_lambda_powertools.utilities.feature_flags import FeatureFlags, AppConfigStore

app_config = AppConfigStore(
    environment="test",
    application="powertools",
    name="test_conf_name",
    max_age=300,
    envelope="features"
)

feature_flags: FeatureFlags = FeatureFlags(store=app_config)

Parameters

store : StoreProvider

Store to use to fetch feature flag schema configuration.

logger : A logging object

Used to log messages. If None is supplied, one will be created.

Expand source code

class FeatureFlags:
    def __init__(self, store: StoreProvider, logger: logging.Logger | Logger | None = None):
        """Evaluates whether feature flags should be enabled based on a given context.

        It uses the provided store to fetch feature flag rules before evaluating them.

        Examples
        --------

        ```python
        from aws_lambda_powertools.utilities.feature_flags import FeatureFlags, AppConfigStore

        app_config = AppConfigStore(
            environment="test",
            application="powertools",
            name="test_conf_name",
            max_age=300,
            envelope="features"
        )

        feature_flags: FeatureFlags = FeatureFlags(store=app_config)
        ```

        Parameters
        ----------
        store: StoreProvider
            Store to use to fetch feature flag schema configuration.
        logger: A logging object
            Used to log messages. If None is supplied, one will be created.
        """
        self.store = store
        self.logger = logger or logging.getLogger(__name__)
        self._exception_handlers: dict[Exception, Callable] = {}

    def _match_by_action(self, action: str, condition_value: Any, context_value: Any) -> bool:
        try:
            func = RULE_ACTION_MAPPING.get(action, lambda a, b: False)
            return func(context_value, condition_value)
        except Exception as exc:
            self.logger.debug(f"caught exception while matching action: action={action}, exception={str(exc)}")

            handler = self._lookup_exception_handler(exc)
            if handler:
                self.logger.debug("Exception handler found! Delegating response.")
                return handler(exc)

            return False

    def _evaluate_conditions(
        self,
        rule_name: str,
        feature_name: str,
        rule: dict[str, Any],
        context: dict[str, Any],
    ) -> bool:
        """Evaluates whether context matches conditions, return False otherwise"""
        rule_match_value = rule.get(schema.RULE_MATCH_VALUE)
        conditions = cast(List[dict], rule.get(schema.CONDITIONS_KEY))

        if not conditions:
            self.logger.debug(
                f"rule did not match, no conditions to match, rule_name={rule_name}, rule_value={rule_match_value}, "
                f"name={feature_name} ",
            )
            return False

        for condition in conditions:
            context_value = context.get(condition.get(schema.CONDITION_KEY, ""))
            cond_action = condition.get(schema.CONDITION_ACTION, "")
            cond_value = condition.get(schema.CONDITION_VALUE)

            # time based rule actions have no user context. the context is the condition key
            if cond_action in (
                schema.RuleAction.SCHEDULE_BETWEEN_TIME_RANGE.value,
                schema.RuleAction.SCHEDULE_BETWEEN_DATETIME_RANGE.value,
                schema.RuleAction.SCHEDULE_BETWEEN_DAYS_OF_WEEK.value,
            ):
                context_value = condition.get(schema.CONDITION_KEY)  # e.g., CURRENT_TIME

            if not self._match_by_action(action=cond_action, condition_value=cond_value, context_value=context_value):
                self.logger.debug(
                    f"rule did not match action, rule_name={rule_name}, rule_value={rule_match_value}, "
                    f"name={feature_name}, context_value={str(context_value)} ",
                )
                return False  # context doesn't match condition

        self.logger.debug(f"rule matched, rule_name={rule_name}, rule_value={rule_match_value}, name={feature_name}")
        return True

    def _evaluate_rules(
        self,
        *,
        feature_name: str,
        context: dict[str, Any],
        feat_default: Any,
        rules: dict[str, Any],
        boolean_feature: bool,
    ) -> bool:
        """Evaluates whether context matches rules and conditions, otherwise return feature default"""
        for rule_name, rule in rules.items():
            rule_match_value = rule.get(schema.RULE_MATCH_VALUE)

            # Context might contain PII data; do not log its value
            self.logger.debug(
                f"Evaluating rule matching, rule={rule_name}, feature={feature_name}, default={str(feat_default)}, boolean_feature={boolean_feature}",  # noqa: E501
            )
            if self._evaluate_conditions(rule_name=rule_name, feature_name=feature_name, rule=rule, context=context):
                # Maintenance: Revisit before going GA.
                return bool(rule_match_value) if boolean_feature else rule_match_value

        # no rule matched, return default value of feature
        self.logger.debug(
            f"no rule matched, returning feature default, default={str(feat_default)}, name={feature_name}, boolean_feature={boolean_feature}",  # noqa: E501
        )
        return feat_default

    def get_configuration(self) -> dict:
        """Get validated feature flag schema from configured store.

        Largely used to aid testing, since it's called by `evaluate` and `get_enabled_features` methods.

        Raises
        ------
        ConfigurationStoreError
            Any propagated error from store
        SchemaValidationError
            When schema doesn't conform with feature flag schema

        Returns
        ------
        dict[str, dict]
            parsed JSON dictionary

            **Example**

        ```python
        {
            "premium_features": {
                "default": False,
                "rules": {
                    "customer tier equals premium": {
                        "when_match": True,
                        "conditions": [
                            {
                                "action": "EQUALS",
                                "key": "tier",
                                "value": "premium",
                            }
                        ],
                    }
                },
            },
            "feature_two": {
                "default": False
            }
        }
        ```
        """
        # parse result conf as JSON, keep in cache for max age defined in store
        self.logger.debug(f"Fetching schema from registered store, store={self.store}")
        config: dict = self.store.get_configuration()
        validator = schema.SchemaValidator(schema=config, logger=self.logger)
        validator.validate()

        return config

    def evaluate(self, *, name: str, context: dict[str, Any] | None = None, default: JSONType) -> JSONType:
        """Evaluate whether a feature flag should be enabled according to stored schema and input context

        **Logic when evaluating a feature flag**

        1. Feature exists and a rule matches, returns when_match value
        2. Feature exists but has either no rules or no match, return feature default value
        3. Feature doesn't exist in stored schema, encountered an error when fetching -> return default value provided

        ┌────────────────────────┐      ┌────────────────────────┐       ┌────────────────────────┐
        │     Feature flags      │──────▶   Get Configuration    ├───────▶     Evaluate rules     │
        └────────────────────────┘      │                        │       │                        │
                                        │┌──────────────────────┐│       │┌──────────────────────┐│
                                        ││     Fetch schema     ││       ││      Match rule      ││
                                        │└───────────┬──────────┘│       │└───────────┬──────────┘│
                                        │            │           │       │            │           │
                                        │┌───────────▼──────────┐│       │┌───────────▼──────────┐│
                                        ││     Cache schema     ││       ││   Match condition    ││
                                        │└───────────┬──────────┘│       │└───────────┬──────────┘│
                                        │            │           │       │            │           │
                                        │┌───────────▼──────────┐│       │┌───────────▼──────────┐│
                                        ││   Validate schema    ││       ││     Match action     ││
                                        │└──────────────────────┘│       │└──────────────────────┘│
                                        └────────────────────────┘       └────────────────────────┘

        Parameters
        ----------
        name: str
            feature name to evaluate
        context: dict[str, Any] | None
            Attributes that should be evaluated against the stored schema.

            for example: `{"tenant_id": "X", "username": "Y", "region": "Z"}`
        default: JSONType
            default value if feature flag doesn't exist in the schema,
            or there has been an error when fetching the configuration from the store
            Can be boolean or any JSON values for non-boolean features.


        Examples
        --------

        ```python
        from aws_lambda_powertools.utilities.feature_flags import AppConfigStore, FeatureFlags
        from aws_lambda_powertools.utilities.typing import LambdaContext

        app_config = AppConfigStore(environment="dev", application="product-catalogue", name="features")

        feature_flags = FeatureFlags(store=app_config)


        def lambda_handler(event: dict, context: LambdaContext):
            # Get customer's tier from incoming request
            ctx = {"tier": event.get("tier", "standard")}

            # Evaluate whether customer's tier has access to premium features
            # based on `has_premium_features` rules
            has_premium_features: bool = feature_flags.evaluate(name="premium_features", context=ctx, default=False)
            if has_premium_features:
                # enable premium features
                ...
        ```

        Returns
        ------
        JSONType
            whether feature should be enabled (bool flags) or JSON value when non-bool feature matches

        Raises
        ------
        SchemaValidationError
            When schema doesn't conform with feature flag schema
        """
        if context is None:
            context = {}

        try:
            features = self.get_configuration()
        except ConfigurationStoreError as err:
            self.logger.debug(f"Failed to fetch feature flags from store, returning default provided, reason={err}")
            return default

        feature = features.get(name)
        if feature is None:
            self.logger.debug(f"Feature not found; returning default provided, name={name}, default={default}")
            return default

        rules = feature.get(schema.RULES_KEY)
        feat_default = feature.get(schema.FEATURE_DEFAULT_VAL_KEY)
        # Maintenance: Revisit before going GA. We might to simplify customers on-boarding by not requiring it
        # for non-boolean flags. It'll need minor implementation changes, docs changes, and maybe refactor
        # get_enabled_features. We can minimize breaking change, despite Beta label, by having a new
        # method `get_matching_features` returning dict[feature_name, feature_value]
        boolean_feature = feature.get(
            schema.FEATURE_DEFAULT_VAL_TYPE_KEY,
            True,
        )  # backwards compatibility, assume feature flag
        if not rules:
            self.logger.debug(
                f"no rules found, returning feature default, name={name}, default={str(feat_default)}, boolean_feature={boolean_feature}",  # noqa: E501
            )
            # Maintenance: Revisit before going GA. We might to simplify customers on-boarding by not requiring it
            # for non-boolean flags.
            return bool(feat_default) if boolean_feature else feat_default

        self.logger.debug(
            f"looking for rule match, name={name}, default={str(feat_default)}, boolean_feature={boolean_feature}",  # noqa: E501
        )
        return self._evaluate_rules(
            feature_name=name,
            context=context,
            feat_default=feat_default,
            rules=rules,
            boolean_feature=boolean_feature,
        )

    def get_enabled_features(self, *, context: dict[str, Any] | None = None) -> list[str]:
        """Get all enabled feature flags while also taking into account context
        (when a feature has defined rules)

        Parameters
        ----------
        context: dict[str, Any] | None
            dict of attributes that you would like to match the rules
            against, can be `{'tenant_id: 'X', 'username':' 'Y', 'region': 'Z'}` etc.

        Returns
        ----------
        list[str]
            list of all feature names that either matches context or have True as default

            **Example**

        ```python
        ["premium_features", "my_feature_two", "always_true_feature"]
        ```

        Raises
        ------
        SchemaValidationError
            When schema doesn't conform with feature flag schema
        """
        if context is None:
            context = {}

        features_enabled: list[str] = []

        try:
            features: dict[str, Any] = self.get_configuration()
        except ConfigurationStoreError as err:
            self.logger.debug(f"Failed to fetch feature flags from store, returning empty list, reason={err}")
            return features_enabled

        self.logger.debug("Evaluating all features")
        for name, feature in features.items():
            rules = feature.get(schema.RULES_KEY, {})
            feature_default_value = feature.get(schema.FEATURE_DEFAULT_VAL_KEY)
            boolean_feature = feature.get(
                schema.FEATURE_DEFAULT_VAL_TYPE_KEY,
                True,
            )  # backwards compatibility, assume feature flag

            if feature_default_value and not rules:
                self.logger.debug(f"feature is enabled by default and has no defined rules, name={name}")
                features_enabled.append(name)
            elif self._evaluate_rules(
                feature_name=name,
                context=context,
                feat_default=feature_default_value,
                rules=rules,
                boolean_feature=boolean_feature,
            ):
                self.logger.debug(f"feature's calculated value is True, name={name}")
                features_enabled.append(name)

        return features_enabled

    def validation_exception_handler(self, exc_class: Exception | list[Exception]):
        """Registers function to handle unexpected validation exceptions when evaluating flags.

        It does not override the function of a default flag value in case of network and IAM permissions.
        For example, you won't be able to catch ConfigurationStoreError exception.

        Parameters
        ----------
        exc_class : Exception | list[Exception]
            One or more exceptions to catch

        Examples
        --------

        ```python
        feature_flags = FeatureFlags(store=app_config)

        @feature_flags.validation_exception_handler(Exception)  # any exception
        def catch_exception(exc):
            raise TypeError("re-raised") from exc
        ```
        """

        def register_exception_handler(func: Callable[P, T]) -> Callable[P, T]:
            if isinstance(exc_class, list):
                for exp in exc_class:
                    self._exception_handlers[exp] = func
            else:
                self._exception_handlers[exc_class] = func

            return func

        return register_exception_handler

    def _lookup_exception_handler(self, exc: BaseException) -> Callable | None:
        # Use "Method Resolution Order" to allow for matching against a base class
        # of an exception
        for cls in type(exc).__mro__:
            if cls in self._exception_handlers:
                return self._exception_handlers[cls]  # type: ignore[index] # index is correct
        return None

Methods

def evaluate(self, *, name: str, context: dict[str, Any] | None = None, default: JSONType)

Evaluate whether a feature flag should be enabled according to stored schema and input context

Logic when evaluating a feature flag

1. Feature exists and a rule matches, returns when_match value
2. Feature exists but has either no rules or no match, return feature default value
3. Feature doesn't exist in stored schema, encountered an error when fetching -> return default value provided

┌────────────────────────┐ ┌────────────────────────┐ ┌────────────────────────┐ │ Feature flags │──────▶ Get Configuration ├───────▶ Evaluate rules │ └────────────────────────┘ │ │ │ │ │┌──────────────────────┐│ │┌──────────────────────┐│ ││ Fetch schema ││ ││ Match rule ││ │└───────────┬──────────┘│ │└───────────┬──────────┘│ │ │ │ │ │ │ │┌───────────▼──────────┐│ │┌───────────▼──────────┐│ ││ Cache schema ││ ││ Match condition ││ │└───────────┬──────────┘│ │└───────────┬──────────┘│ │ │ │ │ │ │ │┌───────────▼──────────┐│ │┌───────────▼──────────┐│ ││ Validate schema ││ ││ Match action ││ │└──────────────────────┘│ │└──────────────────────┘│ └────────────────────────┘ └────────────────────────┘

Parameters

name : str

feature name to evaluate

context : dict[str, Any] | None

Attributes that should be evaluated against the stored schema.

for example: {"tenant_id": "X", "username": "Y", "region": "Z"}

default : JSONType

default value if feature flag doesn't exist in the schema, or there has been an error when fetching the configuration from the store Can be boolean or any JSON values for non-boolean features.

Examples

from aws_lambda_powertools.utilities.feature_flags import AppConfigStore, FeatureFlags
from aws_lambda_powertools.utilities.typing import LambdaContext

app_config = AppConfigStore(environment="dev", application="product-catalogue", name="features")

feature_flags = FeatureFlags(store=app_config)


def lambda_handler(event: dict, context: LambdaContext):
    # Get customer's tier from incoming request
    ctx = {"tier": event.get("tier", "standard")}

    # Evaluate whether customer's tier has access to premium features
    # based on `has_premium_features` rules
    has_premium_features: bool = feature_flags.evaluate(name="premium_features", context=ctx, default=False)
    if has_premium_features:
        # enable premium features
        ...

Returns

JSONType

whether feature should be enabled (bool flags) or JSON value when non-bool feature matches

Raises

SchemaValidationError

When schema doesn't conform with feature flag schema

def get_configuration(self) ‑> dict

Get validated feature flag schema from configured store.

Largely used to aid testing, since it's called by evaluate and get_enabled_features methods.

Raises

ConfigurationStoreError

Any propagated error from store

SchemaValidationError

When schema doesn't conform with feature flag schema

Returns

dict[str, dict]

parsed JSON dictionary

Example

{
    "premium_features": {
        "default": False,
        "rules": {
            "customer tier equals premium": {
                "when_match": True,
                "conditions": [
                    {
                        "action": "EQUALS",
                        "key": "tier",
                        "value": "premium",
                    }
                ],
            }
        },
    },
    "feature_two": {
        "default": False
    }
}

def get_enabled_features(self, *, context: dict[str, Any] | None = None) ‑> list[str]

Get all enabled feature flags while also taking into account context (when a feature has defined rules)

Parameters

context : dict[str, Any] | None

dict of attributes that you would like to match the rules against, can be {'tenant_id: 'X', 'username':' 'Y', 'region': 'Z'} etc.

Returns

list[str]

list of all feature names that either matches context or have True as default

Example

["premium_features", "my_feature_two", "always_true_feature"]

Raises

SchemaValidationError

When schema doesn't conform with feature flag schema

def validation_exception_handler(self, exc_class: Exception | list[Exception])

Registers function to handle unexpected validation exceptions when evaluating flags.

It does not override the function of a default flag value in case of network and IAM permissions. For example, you won't be able to catch ConfigurationStoreError exception.

Parameters

exc_class : Exception | list[Exception]

One or more exceptions to catch

Examples

feature_flags = FeatureFlags(store=app_config)

@feature_flags.validation_exception_handler(Exception)  # any exception
def catch_exception(exc):
    raise TypeError("re-raised") from exc