evaluation

README

LaunchDarkly Go SDK Evaluation Engine

Overview

This repository contains the internal feature flag evaluation logic and data model used by the LaunchDarkly Go SDK. It is packaged separately because it is also used by internal LaunchDarkly components. Applications using the LaunchDarkly Go SDK should not need to reference this package directly.

Supported Go versions

This version of the project requires a Go version of 1.18 or higher.

Integration with easyjson

By default, go-server-sdk-evaluation-private uses LaunchDarkly's open-source JSON library go-jsonstream to convert its data model types like FeatureFlag to and from JSON; this is considerably faster than Go's built-in encoding/json and does not depend on any third-party code. However, it can optionally integrate with the third-party library easyjson, which may be even faster in some cases, without requiring any changes in your code. To enable this, set the build tag launchdarkly_easyjson when you run go build, which both switches go-jsonstream to use easyjson internally and also generates MarshalEasyJSON/UnmarshalEasyJSON methods for each JSON-serializable type. The easyjson library is still under development and has some potential compatibility issues; see its documentation for more details.

If you do not set the launchdarkly_easyjson build tag, go-server-sdk-evaluation does not use any code from easyjson.

Learn more

Check out our documentation for in-depth instructions on configuring and using LaunchDarkly. You can also head straight to the complete reference guide for the Go SDK, or the generated API documentation for this project.

Contributing

We encourage pull requests and other contributions from the community. Check out our contributing guidelines for instructions on how to contribute to this SDK.

About LaunchDarkly

• LaunchDarkly is a continuous delivery platform that provides feature flags as a service and allows developers to iterate quickly and safely. We allow you to easily flag your features and manage them from the LaunchDarkly dashboard. With LaunchDarkly, you can:
  • Roll out a new feature to a subset of your users (like a group of users who opt-in to a beta tester group), gathering feedback and bug reports from real-world use cases.
  • Gradually roll out a feature to an increasing percentage of users, and track the effect that the feature has on key metrics (for instance, how likely is a user to complete a purchase if they have feature A versus feature B?).
  • Turn off a feature that you realize is causing performance problems in production, without needing to re-deploy, or even restart the application with a changed configuration file.
  • Grant access to certain features based on user attributes, like payment plan (eg: users on the ‘gold’ plan get access to more features than users in the ‘silver’ plan). Disable parts of your application to facilitate maintenance, without taking everything offline.
• LaunchDarkly provides feature flag SDKs for a wide variety of languages and technologies. Check out our documentation for a complete list.
• Explore LaunchDarkly
  • launchdarkly.com for more information
  • docs.launchdarkly.com for our documentation and SDK reference guides
  • apidocs.launchdarkly.com for our API documentation
  • blog.launchdarkly.com for the latest product updates

Documentation

Overview

Package evaluation contains the LaunchDarkly Go SDK feature flag evaluation engine.

Normal use of the Go SDK does not require referencing this package directly. It is used internally by the SDK, but is published and versioned separately so it can be used in other LaunchDarkly components without making the SDK versioning dependent on these internal APIs.

Index

• type BigSegmentMembership
• type BigSegmentProvider
• type DataProvider
• type Evaluator
  • func NewEvaluator(dataProvider DataProvider) Evaluator
  • func NewEvaluatorWithOptions(dataProvider DataProvider, options ...EvaluatorOption) Evaluator
• type EvaluatorOption
  • func EvaluatorOptionBigSegmentProvider(bigSegmentProvider BigSegmentProvider) EvaluatorOption
  • func EvaluatorOptionEnableSecondaryKey(enable bool) EvaluatorOption
  • func EvaluatorOptionErrorLogger(errorLogger ldlog.BaseLogger) EvaluatorOption
• type PrerequisiteFlagEvent
• type PrerequisiteFlagEventRecorder
• type Result

Types

type BigSegmentMembership

type BigSegmentMembership interface {
	// CheckMembership tests whether the user is explicitly included or explicitly excluded in the
	// specified segment, or neither. The segment is identified by a segmentRef which is not the
	// same as the segment key-- it includes the key but also versioning information that the SDK
	// will provide. The store implementation should not be concerned with the format of this.
	//
	// If the user is explicitly included (regardless of whether the user is also explicitly
	// excluded or not-- that is, inclusion takes priority over exclusion), the method returns an
	// OptionalBool with a true value.
	//
	// If the user is explicitly excluded, and is not explicitly included, the method returns an
	// OptionalBool with a false value.
	//
	// If the user's status in the segment is undefined, the method returns OptionalBool{} with no
	// value (so calling IsDefined() on it will return false).
	CheckMembership(segmentRef string) ldvalue.OptionalBool
}

BigSegmentMembership is the return type of BigSegmentProvider.GetContextMembership(). It is associated with a single context kind and context key, and provides the ability to check whether that context is included in or excluded from any number of big segments.

This is an immutable snapshot of the state for this context at the time GetBigSegmentMembership was called. Calling CheckMembership should not cause the state to be queried again. The object should be safe for concurrent access by multiple goroutines.

This interface also exists in go-server-sdk because it is exposed as part of the public SDK API; users can write their own implementations of SDK components, but we do not want application code to reference go-server-sdk-evaluation symbols directly as part of that, because this library is versioned separately from the SDK. Currently the two interfaces are identical, but it might be that the go-server-sdk-evaluation version would diverge from the go-server-sdk version due to some internal requirements that aren't relevant to users, in which case go-server-sdk would be responsible for bridging the difference.

type BigSegmentProvider

type BigSegmentProvider interface {
	// GetMembership queries a snapshot of the current segment state for a specific context
	// key.
	//
	// The underlying big segment store implementation will use a hash of the context key, rather
	// than the raw key. But computing the hash is the responsibility of the BigSegmentProvider
	// implementation rather than the evaluator, because there may already have a cached result for
	// that user, and we don't want to have to compute a hash repeatedly just to query a cache.
	//
	// Any given big segment is specific to one context kind, so we do not specify a context kind
	// here; it is OK for the membership results to include different context kinds for the same
	// key. That is, if for instance the context {kind: "user", key: "x"} is included in big segment
	// S1, and the context {kind: "org", key: "x"} is included in big segment S2, then the query
	// result for key "x" will show that it is included in both S1 and S2; even though those "x"
	// keys are really for two unrelated context kinds, we will always know which kind we mean if
	// we are specifically checking either S1 or S2, because S1 is defined as only applying to the
	// "user" kind and S2 is defined as only applying to the "org" kind.
	//
	// If the returned BigSegmentMembership is nil, it is treated the same as an implementation
	// whose CheckMembership method always returns an empty value.
	GetMembership(
		contextKey string,
	) (BigSegmentMembership, ldreason.BigSegmentsStatus)
}

BigSegmentProvider is an abstraction for querying membership in big segments. The caller provides an implementation of this interface to NewEvaluatorWithBigSegments.

type DataProvider

type DataProvider interface {
	// GetFeatureFlag attempts to retrieve a feature flag from the data store by key.
	//
	// The evaluator calls this method if a flag contains a prerequisite condition referencing
	// another flag.
	//
	// The method returns nil if the flag was not found. The DataProvider should treat any deleted
	// flag as "not found" even if the data store contains a deleted flag placeholder for it.
	GetFeatureFlag(key string) *ldmodel.FeatureFlag
	// GetSegment attempts to retrieve a user segment from the data store by key.
	//
	// The evaluator calls this method if a clause in a flag rule uses the OperatorSegmentMatch
	// test.
	//
	// The method returns nil if the segment was not found. The DataProvider should treat any deleted
	// segment as "not found" even if the data store contains a deleted segment placeholder for it.
	GetSegment(key string) *ldmodel.Segment
}

DataProvider is an abstraction for querying feature flags and user segments from a data store. The caller provides an implementation of this interface to NewEvaluator.

Flags and segments are returned by reference for efficiency only (on the assumption that the caller already has these objects in memory); the evaluator will never modify their properties.

type Evaluator

type Evaluator interface {
	// Evaluate evaluates a feature flag for the specified context.
	//
	// The flag is passed by reference only for efficiency; the evaluator will never modify any flag
	// properties. Passing a nil flag will result in a panic.
	//
	// The evaluator does not know anything about analytics events; generating any appropriate analytics
	// events is the responsibility of the caller, who can also provide a callback in prerequisiteFlagEventRecorder
	// to be notified if any additional evaluations were done due to prerequisites. The prerequisiteFlagEventRecorder
	// parameter can be nil if you do not need to track prerequisite evaluations.
	Evaluate(
		flag *ldmodel.FeatureFlag,
		context ldcontext.Context,
		prerequisiteFlagEventRecorder PrerequisiteFlagEventRecorder,
	) Result
}

Evaluator is the engine for evaluating feature flags.

func NewEvaluator

func NewEvaluator(dataProvider DataProvider) Evaluator

NewEvaluator creates an Evaluator, specifying a DataProvider that it will use if it needs to query additional feature flags or user segments during an evaluation.

To support big segments, you must use NewEvaluatorWithOptions and EvaluatorOptionBigSegmentProvider.

func NewEvaluatorWithOptions

func NewEvaluatorWithOptions(dataProvider DataProvider, options ...EvaluatorOption) Evaluator

NewEvaluatorWithOptions creates an Evaluator, specifying a DataProvider that it will use if it needs to query additional feature flags or user segments during an evaluation, and also any number of EvaluatorOption modifiers.

type EvaluatorOption

type EvaluatorOption interface {
	// contains filtered or unexported methods
}

EvaluatorOption is an optional parameter for NewEvaluator.

func EvaluatorOptionBigSegmentProvider

func EvaluatorOptionBigSegmentProvider(bigSegmentProvider BigSegmentProvider) EvaluatorOption

EvaluatorOptionBigSegmentProvider is an option for NewEvaluator that specifies a BigSegmentProvider for evaluating big segment membership. If the parameter is nil, it will be treated the same as a BigSegmentProvider that always returns a "store not configured" status.

func EvaluatorOptionEnableSecondaryKey

func EvaluatorOptionEnableSecondaryKey(enable bool) EvaluatorOption

EvaluatorOptionEnableSecondaryKey is an option for NewEvaluator that specifies whether to enable the use of the ldcontext.Secondary meta-attribute in experiments that involve rollouts or experiments. By default, this is not enabled in the current Go SDK and Rust SDK, and the evaluation engines in other server-side SDKs do not recognize the secondary key meta-attribute at all; but the Go and Rust evaluation engines need to be able to recognize it when they are doing evaluations involving old-style user data.

func EvaluatorOptionErrorLogger

func EvaluatorOptionErrorLogger(errorLogger ldlog.BaseLogger) EvaluatorOption

EvaluatorOptionErrorLogger is an option for NewEvaluator that specifies a logger for error reporting. The Evaluator will only log errors for conditions that should not be possible and require investigation, such as a malformed flag or a code path that should not have been reached. If the parameter is nil, no logging is done.

type PrerequisiteFlagEvent

type PrerequisiteFlagEvent struct {
	// TargetFlagKey is the key of the feature flag that had a prerequisite.
	TargetFlagKey string
	// Context is the context that the flag was evaluated for. We pass this back to the caller, even though the caller
	// already passed it to us in the Evaluate parameters, so that the caller can provide a stateless function for
	// PrerequisiteFlagEventRecorder rather than a closure (since closures are less efficient).
	Context ldcontext.Context
	// PrerequisiteFlag is the full configuration of the prerequisite flag. We need to pass the full flag here rather
	// than just the key because the flag's properties (such as TrackEvents) can affect how events are generated.
	// This is passed by reference for efficiency only, and will never be nil; the PrerequisiteFlagEventRecorder
	// must not modify the flag's properties.
	PrerequisiteFlag *ldmodel.FeatureFlag
	// PrerequisiteResult is the result of evaluating the prerequisite flag.
	PrerequisiteResult Result
	// ExcludeFromSummaries determines if the event will be included in summary information.
	ExcludeFromSummaries bool
}

PrerequisiteFlagEvent is the parameter data passed to PrerequisiteFlagEventRecorder.

type PrerequisiteFlagEventRecorder

type PrerequisiteFlagEventRecorder func(PrerequisiteFlagEvent)

PrerequisiteFlagEventRecorder is a function that Evaluator.Evaluate() will call to record the result of a prerequisite flag evaluation.

type Result

type Result struct {
	// Detail contains the evaluation detail fields.
	Detail ldreason.EvaluationDetail

	// IsExperiment is true if this evaluation result was determined by an experiment. Normally if
	// this is true, then Detail.Reason will also communicate that fact, but there are some cases
	// related to the older experimentation model where this field may be true even if Detail.Reason
	// does not say anything special. When the SDK submits evaluation information to the event
	// processor, it should set the RequireReason field in ldevents.FlagEventProperties to this value.
	IsExperiment bool
}

Result encapsulates all information returned by Evaluator.Evaluate.