experiment

Documentation

Overview

Package experiment abtest Experimental diversion related implementation

Package experiment abtest

Package experiment TODO

Index

• Variables
• func IsHitTag(ctx context.Context, tagListGroup []*protoccacheserver.TagList, ...) (bool, error)
• type Experiment
• type Options

Variables

var (
	// Executor abtest Experimental Shunt Logic Actuator
	Executor = &executor{}
)

Functions

func IsHitTag

func IsHitTag(ctx context.Context, tagListGroup []*protoccacheserver.TagList, options *Options) (bool, error)

IsHitTag Whether the tag is hit

Types

type Experiment

type Experiment struct {
	*protoccacheserver.Group // The hit experiment group, Group must be not empty
	IsOverrideList           bool
	HoldoutData              map[string]*Experiment
}

Experiment Experimental information, encapsulates the group in the protocol, and adds status attributes during the diversion process IsOverrideList Whether it is the experimental group hit by the whitelist

type Options

type Options struct {
	// Whether TAB automatically records exposure
	IsExposureLoggingAutomatic bool `json:"ela,omitempty"`
	// Scene ID, used for filtering, returns the hit experiment groups in the lower layers of these scenes,
	// key is sceneID, value is whether it passes, if sceneIDs is empty, all passes
	SceneIDs map[int64]bool `json:"sceneIDs,omitempty"`
	// Layer key, used for filtering, returns the hit experiment groups under these layers,
	// key is layerKey, value is whether it passes, if layerKeys is empty, all passes
	LayerKeys map[string]bool `json:"layerKeys,omitempty"`
	// Experiment key, used for filtering, returns the experimental group hit in this experiment.
	// Since there are multiple experiments in the same layer, if the user hits other experiments in the same layer,
	// the experimental groups of other experiments will not be returned, but the default experiment will be returned.
	ExperimentKeys map[string]bool `json:"experimentKeys,omitempty"`
	// Whether to pre-process the dmp tag. If enabled, if there is a dmp tag under the business,
	// rpc will first access the dmp service to get the hit or not result. Enabled by default
	IsPreparedDMPTag bool `json:"isPreparedDmpTag,omitempty"`
	// The key format is {{reporting ID}}-{{DMP platform ID}}-{{crowd package key}},
	// and the value is whether it is a hit
	DMPTagResult map[string]bool `json:"dmpTagResult,omitempty"`
	// Whether to disable dmp. If disabled, no rpc will be initiated.
	// ab test will offload all local calculations and will not hit dmp by default
	IsDisableDMP bool `json:"isDisableDmp,omitempty"`
	// Whitelist, key is the layer, value is the experiment group specified by the user under the layer
	OverrideList map[string]int64 `json:"-"`
	// Attribute tag information owned by unitID
	AttributeTag map[string][]string `json:"-"`
	// unitID passed in by the user
	UnitID string `json:"unitId,omitempty"`
	// The ID used for hashing is usually consistent with the unitID.
	// The same UnitID will be mapped to the same DecisionID, that is,
	// the same UnitID will stably hit the same experimental group
	DecisionID string `json:"decisionId,omitempty"`
	// The newID passed in by the user is usually empty. If the user account system is switching to grayscale,
	// both unitID and newUnitID can be passed in. In abtest diversion,
	// according to the layer properties set by the web platform,
	// the specified unitID or newUnitID is selected for reporting
	NewUnitID string `json:"newUnitId,omitempty"`
	// ID used for hashing, usually consistent with newUnitID.
	// If abtest splits traffic and the layer specifies to use NewUnitID,
	// NewDecisionID will be used as the input of hashing. The same NewUnitID and the same NewDecisionID
	// will stably hit the same experimental group.
	NewDecisionID string `json:"newDecisionId,omitempty"`
	// Cache data snapshot
	Application *cache.Application `json:"-"`
	// The result of the holdout layer hit. If it is nil, it means that it is not held out.
	HoldoutLayerResult map[string]*Experiment `json:"-"`
}

Options abtest experiment diversion related options, life cycle for each abtest diversion session