Skip to main content

Build Your Own SDK

Latest spec version: 0.7.0 View Changelog

This guide is meant for library authors looking to build a GrowthBook SDK in a currently unsupported language.

All libraries should follow this specification as closely as the language permits to maintain consistency and make updates and maintenance easier.

At the end of this guide is an extensive test suite with over 400 language-agnostic unit tests in JSON format. All SDKs must pass 100% of these test cases. It's recommended to add additional manual unit tests as needed on top of these.

Data structures

Here are a number of important data structures in GrowthBook SDKs, listed alphabetically.

Attributes

Attributes are an arbitrary JSON object containing user and request attributes. Here's an example:

{
"id": "123",
"anonId": "abcdef",
"company": "growthbook",
"url": "/pricing",
"country": "US",
"browser": "firefox",
"age": 25,
"beta": true,
"account": {
"plan": "team",
"seats": 10
}
}

BucketRange

A tuple that describes a range of the numberline between 0 and 1.

The tuple has 2 parts, both floats - the start of the range and the end. For example:

[0.3, 0.7];

Condition

A Condition is evaluated against Attributes and used to target features/experiments to specific users.

The syntax is inspired by MongoDB queries. Here is an example:

{
"country": "US",
"browser": {
"$in": ["firefox", "chrome"]
},
"email": {
"$not": {
"$regex": "@gmail.com$"
}
}
}

ParentCondition

A ParentCondition defines a prerequisite. It consists of a parent feature's id (string), a condition (Condition), and an optional gate (boolean) flag.

Instead of evaluating against attributes, the condition evaluates against the returned value of the parent feature. The condition will always reference a "value" property. Here is an example of a gating prerequisite where the parent feature must be on (true)

{
"id": "parent-feature",
"condition": {
"value": {
"$eq": true
}
},
"gate": true
}

Context

Context object passed into the GrowthBook constructor. Has a number of optional properties:

  • enabled (boolean) - Switch to globally disable all experiments. Default true.
  • apiHost (string) - The GrowthBook API Host. Optional
  • clientKey (string) - The key used to fetch features from the GrowthBook API. Optional
  • decryptionKey (string) - The key used to decrypt encrypted features from the API. Optional
  • attributes (Attributes) - Map of user attributes that are used to assign variations
  • url (string) - The URL of the current page
  • features (FeatureMap) - Feature definitions (usually pulled from an API or cache)
  • forcedVariations (ForcedVariationsMap) - Force specific experiments to always assign a specific variation (used for QA)
  • qaMode (boolean) - If true, random assignment is disabled and only explicitly forced variations are used.
  • trackingCallback (TrackingCallback) - A function that takes experiment and result as arguments.

Experiment

Defines a single Experiment. Has a number of properties:

  • key (string) - The globally unique identifier for the experiment
  • variations (any[]) - The different variations to choose between
  • weights (float[]) - How to weight traffic between variations. Must add to 1.
  • active (boolean) - If set to false, always return the control (first variation)
  • coverage (float) - What percent of users should be included in the experiment (between 0 and 1, inclusive)
  • ranges (BucketRange[]) - Array of ranges, one per variation
  • condition (Condition) - Optional targeting condition
  • namespace (Namespace) - Adds the experiment to a namespace
  • force (integer) - All users included in the experiment will be forced into the specific variation index
  • hashAttribute (string) - What user attribute should be used to assign variations (defaults to id)
  • fallbackAttribute (string) - When using sticky bucketing, can be used as a fallback to assign variations
  • hashVersion (integer) - The hash version to use (default to 1)
  • meta (VariationMeta[]) - Meta info about the variations
  • filters (Filter[]) - Array of filters to apply
  • seed (string) - The hash seed to use
  • name (string) - Human-readable name for the experiment
  • phase (string) - Id of the current experiment phase
  • disableStickyBucketing (boolean) - If true, sticky bucketing will be disabled for this experiment. (Note: sticky bucketing is only available if a StickyBucketingService is provided in the Context)
  • bucketVersion (integer) - An sticky bucket version number that can be used to force a re-bucketing of users (default to 0)
  • minBucketVersion (integer) - Any users with a sticky bucket version less than this will be excluded from the experiment

The only required properties are key and variations. Everything else is optional.

ExperimentResult

The result of running an Experiment given a specific Context

  • inExperiment (boolean) - Whether or not the user is part of the experiment
  • variationId (int) - The array index of the assigned variation
  • value (any) - The array value of the assigned variation
  • hashUsed (boolean) - If a hash was used to assign a variation
  • hashAttribute (string) - The user attribute used to assign a variation
  • hashValue (string) - The value of that attribute
  • featureId (string or null) - The id of the feature (if any) that the experiment came from
  • key (string) - The unique key for the assigned variation
  • bucket (float) - The hash value used to assign a variation (float from 0 to 1)
  • name (string or null) - The human-readable name of the assigned variation
  • passthrough (boolean) - Used for holdout groups
  • stickyBucketUsed (boolean) - If sticky bucketing was used to assign a variation

The variationId and value should always be set, even when inExperiment is false.

The hashAttribute and hashValue should always be set, even when hashUsed is false.

The key should always be set, even if experiment.meta is not defined or incomplete. In that case, convert the variation's array index to a string (e.g. 0 -> "0") and use that as the key instead.

Feature

A Feature object consists of a default value plus rules that can override the default.

  • defaultValue (any) - The default value (should use null if not specified)
  • rules (FeatureRule[]) - Array of FeatureRule objects that determine when and how the defaultValue gets overridden

FeatureMap

A hash or map of Feature objects. Keys are string ids for the features. Values are Feature objects. For example:

{
"feature-1": {
"defaultValue": false
},
"my_other_feature": {
"defaultValue": 1,
"rules": [
{
"force": 2
}
]
}
}

FeatureResult

The result of evaluating a Feature. Has a number of properties:

  • value (any) - The assigned value of the feature
  • on (boolean) - The assigned value cast to a boolean
  • off (boolean) - The assigned value cast to a boolean and then negated
  • source (enum) - One of "unknownFeature", "defaultValue", "force", or "experiment"
  • experiment (Experiment or null) - When source is "experiment", this will be an Experiment object
  • experimentResult (ExperimentResult or null) - When source is "experiment", this will be an ExperimentResult object

FeatureRule

Overrides the defaultValue of a Feature. Has a number of optional properties

  • id (string) - Optional rule id, reserved for future use
  • condition (Condition) - Optional targeting condition
  • parentConditions (ParentCondition[]) - Each item defines a prerequisite where a condition must evaluate against a parent feature's value (identified by id). If gate is true, then this is a blocking feature-level prerequisite; otherwise it applies to the current rule only.
  • coverage (float) - What percent of users should be included in the experiment (between 0 and 1, inclusive)
  • force (any) - Immediately force a specific value (ignore every other option besides condition and coverage)
  • variations (any[]) - Run an experiment (A/B test) and randomly choose between these variations
  • key (string) - The globally unique tracking key for the experiment (default to the feature key)
  • weights (float[]) - How to weight traffic between variations. Must add to 1.
  • namespace (Namespace) - Adds the experiment to a namespace
  • hashAttribute (string) - What user attribute should be used to assign variations (defaults to id)
  • hashVersion (integer) - The hash version to use (default to 1)
  • range (BucketRange) - A more precise version of coverage
  • ranges (BucketRange[]) - Ranges for experiment variations
  • meta (VariationMeta[]) - Meta info about the experiment variations
  • filters (Filter[]) - Array of filters to apply to the rule
  • seed (string) - Seed to use for hashing
  • name (string) - Human-readable name for the experiment
  • phase (string) - The phase id of the experiment
  • tracks (TrackData[]) - Array of tracking calls to fire

Filter

Object used for mutual exclusion and filtering users out of experiments based on random hashes. Has the following properties:

  • seed (string) - The seed used in the hash
  • ranges (BucketRange[]) - Array of ranges that are included
  • hashVersion (integer) - The hash version to use (default to 2)
  • attribute (string, optional) - The attribute to use (default to "id")

ForcedVariationsMap

A hash or map that forces an Experiment to always assign a specific variation. Useful for QA.

Keys are the experiment key, values are the array index of the variation. For example:

{
"my-test": 0,
"other-test": 1
}

Namespace

A tuple that specifies what part of a namespace an experiment includes. If two experiments are in the same namespace and their ranges don't overlap, they wil be mutually exclusive.

The tuple has 3 parts:

  1. The namespace id (string)
  2. The beginning of the range (float, between 0 and 1)
  3. The end of the range (float, between 0 and 1)

For example:

["namespace1", 0, 0.5];

TrackingCallback

A callback function that is executed every time a user is included in an Experiment. Here's an example:

function track(experiment, result) {
analytics.track("Experiment Viewed", {
experimentId: experiment.key,
variationId: result.variationId,
});
}

TrackData

Used for remote feature evaluation to trigger the TrackingCallback. An object with 2 properties:

  • experiment - Experiment
  • result - ExperimentResult

VariationMeta

Meta info about an experiment variation. Has the following properties:

  • key (string, optional) - A unique key for this variation
  • name (string, optional) - A human-readable name for this variation
  • passthrough (boolean, optional) - Used to implement holdout groups

Helper Functions

There are some helper functions which are used a few times throughout the SDK.

hash(seed: string, value: string, version: integer): float|null

Hashes a string to a float between 0 and 1.

Uses the simple Fowler–Noll–Vo algorithm, specifically fnv32a. An implementation of this is available in most languages already, and if not it's only a few lines of code to implement yourself. Fnv32a returns an integer, so we convert that to a float using a modulus.

The original hash version (1) had a flaw that caused bias when running experiments in parallel.

// New hashing algorithm
if (version === 2) {
n = fnv32a(fnv32a(seed + value) + "");
return (n % 10000) / 10000;
}
// Original hashing algorithm (with a bias flaw)
else if (version === 1) {
n = fnv32a(value + seed);
return (n % 1000) / 1000;
}

return null;

Note: It's important to use the exact hashing algorithms outlined here so all SDKs behave identically.

inRange(n: float, range: BucketRange): boolean

Determines if a number n is within the provided range.

return n >= range[0] && n < range[1]>;

inNamespace(userId: string, namespace: Namespace): boolean

This checks if a userId is within an experiment namespace or not.

The namespace argument is a tuple with 3 parts: id (string), start (float), and end (float).

  1. Hash the userId and namespace name with two underscores as a delimiter
    n = hash("__" + namespace[0], userId, 1);
  2. Return if hash is greater than (inclusive) the namespace start and less than (exclusive) the namespace end:
    return n >= namespace[1] && n < namespace[2];

getEqualWeights(numVariations: integer): float[]

Returns an array of floats with numVariations items that are all equal and sum to 1. For example, getEqualWeights(2) would return [0.5, 0.5].

It's ok if the sum is slightly off due to rounding. So a sum of 0.9999999 is fine for example.

  1. If numVariations is less than 1, return empty array
  2. Create array with a length of numVariations
  3. Fill the array with 1.0/numVariations and return

getBucketRanges(numVariations: integer, coverage: float, weights: float[]): BucketRange[]

This converts and experiment's coverage and variation weights into an array of bucket ranges.

numVariations is an integer, coverage is a float, and weights is an array of floats.

  1. Clamp the value of coverage to between 0 and 1 inclusive.

    if (coverage < 0) coverage = 0;
    if (coverage > 1) coverage = 1;
  2. Default to equal weights if the weights don't match the number of variations.

    if (weights.length != numVariations) {
    weights = getEqualWeights(numVariations);
    }
  3. Default to equal weights if the sum is not equal 1 (or close enough when rounding errors are factored in):

    if (sum(weights) < 0.99 || sum(weights) > 1.01) {
    weights = getEqualWeights(numVariations);
    }
  4. Convert weights to ranges and return

    cumulative = 0;
    ranges = [];

    for (w in weights) {
    start = cumulative;
    cumulative += w;
    ranges.push([start, start + coverage * w]);
    }

    return ranges;

Some examples:

  • getBucketRanges(2, 1, [0.5, 0.5]) -> [[0, 0.5], [0.5, 1]]
  • getBucketRanges(2, 0.5, [0.4, 0.6]) -> [[0, 0.2], [0.4, 0.7]]

chooseVariation(n: float, ranges: BucketRange[]): integer

Given a hash and bucket ranges, assign one of the bucket ranges.

  1. Loop through ranges
    1. If n is within the range, return the range index
      if (inRange(n, ranges[i])) {
      return i;
      }
  2. Return -1 if it makes it through the whole ranges array without returning

If multiple ranges match, return the first matching one.

getQueryStringOverride(id: string, url: string, numVariations: integer): null|integer

This checks if an experiment variation is being forced via a URL query string. This may not be applicable for all SDKs (e.g. mobile).

As an example, if the id is my-test and url is http://localhost/?my-test=1, you would return 1.

If possible, you should use a proper URL parsing library vs relying on simple regexes.

Return null if any of these are true:

  • There is no querystring
  • The id is not a key in the querystring
  • The variation is not an integer
  • The variation is less than 0 or greater than or equal to numVariations

decrypt(encryptedString: string, decryptionKey: string): string

This decrypts a string using the AES-CBC 128KB algorithm. This is used if the GrowthBook App is configured to encrypt feature flag definitions.

Here's an example in PHP:

function decrypt(string $encryptedString, string $decryptionKey) {
// Split the string into two parts, delimited by "."
list($iv, $cipherText) = explode(".", $encryptedString, 2);

// The Initialization Vector (iv) is base64 encoded
$iv = base64_decode($iv);

// Decrypt using the AES-CBC 128kb algorithm
// Will throw an Exception if unable to decrypt
return openssl_decrypt($cipherText, "aes-128-cbc", $decryptionKey, 0, $iv);
}

The return value will be a JSON-encoded string. If an error occurs, you can throw an exception (or whatever is typically used for error handling).

Evaluating Conditions

In addition to the helper functions above, there are a number of methods related to evaluating targeting conditions.

There is only one public method evalCondition and everything else is a private helper function.

public evalCondition(attributes: Attributes, condition: Condition): boolean

This is the main function used to evaluate a condition. It loops through the condition key/value pairs and checks each entry:

  1. If condition key is $or, check if evalOr(attributes, condition["$or"]) is false. If so, break out of the loop and return false
  2. If condition key is $nor, check if !evalOr(attributes, condition["$nor"]) is false. If so, break out of the loop and return false
  3. If condition key is $and, check if evalAnd(attributes, condition["$and"]) is false. If so, break out of the loop and return false
  4. If condition key is $not, check if !evalCondition(attributes, condition["$not"]) is false. If so, break out of the loop and return false
  5. Otherwise, check if evalConditionValue(value, getPath(attributes, key)) is false. If so, break out of the loop and return false

If none of the entries failed their checks, evalCondition returns true

private evalOr(attributes: Attributes, conditions: Condition[]): boolean

conditions is an array of Condition objects

  1. If conditions is empty, return true
  2. Loop through conditions
    1. If evalCondition(attributes, conditions[i]) is true, break out of the loop and return true
  3. Return false

private evalAnd(attributes: Attributes, conditions: Condition[]): boolean

conditions is an array of Condition objects

  1. Loop through conditions
    1. If evalCondition(attributes, conditions[i]) is false, break out of the loop and return false
  2. Return true

private isOperatorObject(obj): boolean

This accepts a parsed JSON object as input and returns true if every key in the object starts with $.

  • {"$gt": 1} -> true
  • {"$gt": 1, "$lt": 10} -> true
  • {"foo": "bar"} -> false
  • {"$gt": 1, "foo": "bar"} -> false

If the object is empty and has no keys, this should also return true.

private getType(attributeValue): string

This returns the data type of the passed in argument.

The valid types to return are:

  • string
  • number
  • boolean
  • array
  • object
  • null
  • undefined
  • unknown

The difference between null and undefined can be illustrated as follows:

obj = JSON.parse('{"foo": null}');

getType(obj["foo"]); // null

getType(obj["bar"]); // undefined

The value unknown is there just in case you can't figure out the data type for whatever reason. It will never be used in most implementations.

private getPath(attributes: Attributes, path: string): any

Given attributes and a dot-separated path string, return the value at that path (or null/undefined if the path doesn't exist)

Given the input:

{
"name": "john",
"job": {
"title": "developer"
}
}

It should return:

  • getPath(input, "name") -> "john"
  • getPath(input, "job.title") -> "developer"
  • getPath(input, "job.company") -> null or undefined

private evalConditionValue(conditionValue, attributeValue): boolean

  1. If conditionValue is an object and isOperatorObject(conditionValue) is true
    1. Loop over each key/value pair
      1. If evalOperatorCondition(key, attributeValue, value) is false, return false
    2. Return true
  2. Else, do a deep comparison between attributeValue and conditionValue. Return true if equal, false if not.

private elemMatch(conditionValue, attributeValue): boolean

This checks if attributeValue is an array, and if so at least one of the array items must match the condition

  1. If attributeValue is not an array, return false
  2. Loop through items in attributeValue
    1. If isOperatorObject(conditionValue)
      1. If evalConditionValue(conditionValue, item), break out of loop and return true
    2. Else if evalCondition(item, conditionValue), break out of loop and return true
  3. Return false

private paddedVersionString(input): string

This function can be used to help with the evaluation of the version string comparsion.

There are 6 operators that are used for comparing version strings, e.g. v1.2.3 or 1.2.3:

ConditionComparisonDescription
$veq==Versions are equal
$vne!=Versions are not equal
$vlt<The first version is lesser than the second version
$vlte<=The first version is lesser than or equal to the second version
$vgt>The first version is greater than the second version
$vgte>=The first version is greater than or equal to the second version

Rules:

  • Segments are separated by . and - characters
  • Segments should be compared alphanumerically from the left-most segment
    • Digit-only segments should be left-padded with a space so that they have the same number of characters.
  • A leading v in a version string should be ignored
  • Semantic version syntax used to denote build information (as denoted by a +, e.g. +mybuild) should be ignored for comparisons.

Here's an example:

export function paddedVersionString(input: string): string {
// Remove build info and leading `v` if any
// Split version into parts (both core version numbers and pre-release tags)
// "v1.2.3-rc.1+build123" -> ["1","2","3","rc","1"]
const parts = input.replace(/(^v|\+.*$)/g, "").split(/[-.]/);

// If it's SemVer without a pre-release, add `~` to the end
// ["1","0","0"] -> ["1","0","0","~"]
// "~" is the largest ASCII character, so this will make "1.0.0" greater than "1.0.0-beta" for example
if (parts.length === 3) {
parts.push("~");
}

// Left pad each numeric part with spaces so string comparisons will work ("9">"10", but " 9"<"10")
// Then, join back together into a single string
return parts
.map((v) => (v.match(/^[0-9]+$/) ? v.padStart(5, " ") : v))
.join("-");
}

private isIn(conditionValue, actualValue): boolean

Checks to see if actualValue is in the conditionValue array. This implements the $in and $nin operators.

  1. If actualValue is an array
    1. Return true if the intersection between actualValue and conditionValue has at least 1 element. Otherwise, return false.
  2. Else
    1. Return true if conditionValue contains actualValue. Otherwise, return false.

private evalOperatorCondition(operator, attributeValue, conditionValue)

This function is just a case statement that handles all the possible operators

There are basic comparison operators in the form attributeValue {op} conditionValue:

  • $eq ==
  • $ne != (not equals)
  • $lt <
  • $lte <=
  • $gt >
  • $gte >=
  • $regex ~ (regex match)

There are 3 operators where conditionValue is an array. All of these should return false if conditionValue is not an array for whatever reason.

  • $in
    1. Return isIn(conditionValue, attributeValue)
  • $nin
    1. Return not isIn(conditionValue, attributeValue)
  • $all
    1. If attributeValue is not an array, return false
    2. Loop through conditionValue array
      1. If none of the elements in the attributeValue array pass evalConditionValue(conditionValue[i], attributeValue[j]), return false
    3. Return true

There are 2 operators where attributeValue is an array:

  • $elemMatch
    1. Return elemMatch(conditionValue, attributeValue)
  • $size
    1. If attributeValue is not an array, return false
    2. Return evalConditionValue(conditionValue, attributeValue.length)

There are 3 other operators:

  • $exists
    1. If conditionValue is false, return true if attributeValue is null or undefined
    2. Else, return true if attributeValue is NOT null or undefined
    3. Return false by default
  • $type
    1. Return getType(attributeValue) == conditionValue
  • $not
    1. Return !evalConditionValue(conditionValue, attributeValue)

There are 6 operators that are used for comparing version strings, e.g. v1.2.3 or 1.2.3. See paddedVersionString(input) for details.

If operator doesn't match any of these, return false and potentially log the error for debug purposes.

GrowthBook Class

The GrowthBook class is the main export of the SDK.

constructor

The constructor takes a Context object and stores the properties for later. Nothing else needs to be done during initialization.

This class has a few helper methods as well as 2 main public methods - evalFeature and run.

Getters and Setters

There should be simple getters and setters for a few of the context properties:

  • attributes
  • features
  • forcedVariations
  • url
  • enabled

private getFeatureResult(value, source, experiment, experimentResult): FeatureResult

This is a helper method to create a FeatureResult object. The first two arguments, value, and source are required. The last two, experiment and experimentResult are optional and should default to null.

Besides the passed-in arguments, there are two derived values - on and off, which are just the value cast to booleans.

value can be any JSON type. Only the following values are considered to be "falsy":

  • null
  • false
  • ""
  • 0

Everything else is considered "truthy", including empty arrays and objects.

If value is "truthy", then on should be true and off should be false. If the value is "falsy", then they should take opposite values.

private isFilteredOut(filters: Filters[]): boolean

This is a helper method to evaluate filters for both feature flags and experiments.

  1. Loop through filters array
    1. Get the hashAttribute and hashValue
    hashAttribute = filter.attribute || "id";
    hashValue = context.attributes[hashAttribute] || "";
    1. If hashValue is empty, return true
    2. Determine the bucket for the user
      n = hash(filter.seed, hashValue, filter.hashVersion || 2);
    3. If inRange(n, range) is false for every range in filter.ranges, return true
  2. If you made it through the entire array without returning early, return false now

private isIncludedInRollout(seed: string, hashAttribute: string | null, range: BucketRange | null, coverage: float | null, hashVersion: integer | null): boolean

Determines if the user is part of a gradual feature rollout.

  1. Either coverage or range are required. If both are null, return true immediately

  2. If range is null and coverage is zero, return false immediately. This catches an edge case where the bucket is zero and users are let through when they shouldn't be

  3. Get the hashAttribute and hashValue

    hashAttribute = hashAttribute || "id";
    hashValue = context.attributes[hashAttribute] || "";
  4. If hashValue is empty, return false immediately

  5. Determine the bucket for the user

    n = hash(seed, hashValue, hashVersion || 1)
  6. Check if user is included

    if (range) {
    return inRange(n, range)
    }
    else if (coverage !== null) {
    return n <= coverage
    }

    return true

private getExperimentResult(experiment, variationIndex, hashUsed, featureId, bucket): ExperimentResult

This is a helper method to create an ExperimentResult object. The arguments are:

  • experiment - Experiment object (required)
  • variationIndex - The assigned variation index (optional, default to -1)
  • hashUsed - Whether or not the hash was used to assign a variation (optional, default to false)
  • featureId - The id of the feature (if any) that the experiment came from (optional, default to null)
  • bucket - The hash bucket value for the user. Float from 0 to 1 (optional, default to null)

The method is pretty simple:

  1. Handle case when user is not in the experiment (e.g. variationIndex = -1)

    // By default, assume everyone is in the experiment
    let inExperiment = true;
    // If the variation is invalid, use the baseline and set the inExperiment flag to false
    if (variationIndex < 0 || variationIndex >= experiment.variations.length) {
    variationIndex = 0;
    inExperiment = false;
    }
  2. Get the hashAttribute and hashValue

    hashAttribute = experiment.hashAttribute || "id";
    hashValue = context.attributes[hashAttribute] || "";
  3. Get meta info for the assigned variation (if any)

    meta = experiment.meta ? experiment.meta[variationIndex] : null
  4. Build return object

    res = {
    key: (meta && meta.key) ? meta.key : ("" + variationIndex),
    featureId: featureId,
    inExperiment: inExperiment,
    hashUsed: hashUsed,
    variationId: variationIndex,
    value: experiment.variations[variationIndex],
    hashAttribute: hashAttribute,
    hashvalue: hashValue,
    };
  5. Add optional properties and return

    if (meta && meta.name) res.name = meta.name;
    if (meta && meta.passthrough) res.passthrough = true;
    if (bucket !== null) res.bucket = bucket;

    return res;

public evalFeature(key: string): FeatureResult

The evalFeature method takes a single string argument, which is the unique identifier for the feature and returns a FeatureResult object.

growthbook = new GrowthBook(context);
myFeature = growthbook.evalFeature("my-feature");

There are a few ordered steps to evaluate a feature

  1. If the key doesn't exist in context.features
    1. Return getFeatureResult(null, "unknownFeature")
  2. Loop through the feature rules (if any)
    1. If the rule has parentConditions (prerequisites) defined, loop through each one:
      1. Call evalFeature on the parent condition
        • If a cycle is detected, break out of feature evaluation and return getFeatureResult(null, "cyclicPrerequisite")
      2. Using the evaluated parent's result, create an object
        const evalObj = { "value": parentResult.value }
      3. Evaluate this object against the parentCondition's condition:
        evalCondition(evalObj, parentResult.value);
      4. If any of the parentConditions fail evaluation then:
        • If parentCondition.gate is true (a blocking prerequisite), return getFeatureResult(null, "prerequisite")
        • Otherwise, skip this rule and continue to the next one
    2. If the rule has filters defined
      1. if isFilteredOut(rule.filters), skip this rule and continue to the next one
    3. If the rule has a condition
      1. If evalCondition(context.attributes, rule.condition) is false, skip this rule and continue to the next one
    4. If rule.force is set
      1. If not isIncludedInRollout, skip this rule and continue to the next one
        if (!isIncludedInRollout(
        rule.seed || featureKey,
        rule.hashAttribute,
        rule.range,
        rule.coverage,
        rule.hashVersion
        )) {
        continue;
        }
      2. If rule.tracks is set, fire the TrackingCallback for each element in the rule.tracks array.
      3. Return getFeatureResult(rule.force, "force")
    5. Otherwise, convert the rule to an Experiment object
      const exp = {
      variations: rule.variations,
      key: rule.key || featureKey,
      };
    6. Copy over additional settings from the rule to exp if defined:
      • coverage
      • weights
      • hashAttribute
      • fallbackAttribute
      • disableStickyBucketing
      • bucketVersion
      • minBucketVersion
      • namespace
      • meta
      • ranges
      • name
      • phase
      • seed
      • hashVersion
      • filters
      • condition
    7. Run the experiment
      result = run(exp);
    8. If result.inExperiment is false OR result.passthrough is true, skip this rule and continue to the next one
    9. Otherwise, return
      return getFeatureResult(result.value, "experiment", exp, result);
  3. Return getFeatureResult(feature.defaultValue || null, "defaultValue")

public run(experiment: Experiment): ExperimentResult

The run method takes an Experiment object and returns an ExperimentResult.

There are a bunch of ordered steps to run an experiment:

  1. If experiment.variations has fewer than 2 variations, return getExperimentResult(experiment)
  2. If context.enabled is false, return getExperimentResult(experiment)
  3. If context.url exists
    qsOverride = getQueryStringOverride(experiment.key, context.url);
    if (qsOverride != null) {
    return getExperimentResult(experiment, qsOverride);
    }
  4. Return if forced via context
    if (experiment.key in context.forcedVariations) {
    return getExperimentResult(
    experiment,
    context.forcedVariations[experiment.key]
    );
    }
  5. If experiment.active is set to false, return getExperimentResult(experiment)
  6. Get the user hash value and return if empty
    hashAttribute = experiment.hashAttribute || "id";
    hashValue = context.attributes[hashAttribute] || "";
    if (hashValue == "") {
    if (experiment.fallbackAttribute && context.attributes[experiment.fallbackAttribute]) {
    // check if a fallbackAttribute exists (sticky bucketing)
    hashAttribute = experiment.fallbackAttribute;
    hashValue = context.attributes[hashAttribute];
    } else {
    // no hashAttribute or fallbackAttribute, return
    return getExperimentResult(experiment);
    }
    }
    6.5 If sticky bucketing is permitted, check to see if a sticky bucket value exists. If so, skip steps 7-8.
  7. Apply filters and namespace
    1. If experiment.filters is set
      if (isFilteredOut(experiment.filters)) {
      return getExperimentResult(experiment)
      }
    2. Else if experiment.namespace is set, return if not in range
      if (!inNamespace(hashValue, experiment.namespace)) {
      return getExperimentResult(experiment);
      }
  8. Return if any conditions are not met, return
    1. If experiment.condition is set, return if it evaluates to false
    if (!evalCondition(context.attributes, experiment.condition)) {
    return getExperimentResult(experiment);
    }
    1. If experiment.parentConditions is set (prerequisites), return if any of them evaluate to false. See the corresponding logic in evalFeature for more details. (Note that the gate flag should not be set in an experiment)
    2. Apply any url targeting based on experiment.urlPatterns, return if no match
  9. Choose a variation
    1. If a sticky bucket value exists, use it.
      1. If the found sticky bucket version is blocked (doesn't exceed experiment.minBucketVersion), then skip enrollment
    2. Else, calculate bucket ranges for the variations and choose one
      ranges = experiment.ranges || getBucketRanges(
      experiment.variations.length,
      experiment.converage ?? 1,
      experiment.weights ?? []
      );
      n = hash(
      experiment.seed || experiment.key,
      hashValue,
      experiment.hashVersion || 1
      );
      assigned = chooseVariation(n, ranges);
  10. If assigned == -1, return getExperimentResult(experiment)
  11. If experiment has a forced variation, return
    if ("force" in experiment) {
    return getExperimentResult(experiment, experiment.force);
    }
  12. If context.qaMode, return getExperimentResult(experiment)
  13. Build the result object
    result = getExperimentResult(experiment, assigned, true, n);
    ``
    13.5 If sticky bucketing is permitted, store the sticky bucket value
  14. Fire context.trackingCallback if set and the combination of hashAttribute, hashValue, experiment.key, and variationId has not been tracked before
  15. Return result

Feature helper methods

There are 3 tiny helper methods that wrap evalFeature for a better developer experience:

public isOn(key) {
return this.evalFeature(key).on
}

public isOff(key) {
return this.evalFeature(key).off
}

public getFeatureValue(key, fallback) {
value = this.evalFeature(key).value
return value === null ? fallback : value
}

For strongly typed languages, you can use generics (if supported) for getFeatureValue and coerce the return value to always match the data type of fallback. If generics are not supported, you can use type-specific versions of the function such as getFeatureValueAsString.

Fetching and Caching Features

When the Context contains a clientKey, the SDK should fetch and cache features automatically.

If apiHost is not specified, default to https://cdn.growthbook.io. Make sure to strip and trailing slashes on user-entered hosts (e.g. http://example.com/ becomes http://example.com).

Features should be fetched from {apiHost}/api/features/{clientKey} and all errors should be handled gracefully. A network error while fetching features should never be a fatal error that stops execution.

The API responses should be parsed and cached so future GrowthBook instances with the same clientKey can avoid a duplicate network request. The standard cache TTL to use is 60 seconds.

For best performance, a stale-while-revalidate pattern should be used. If a cache entry is older than the TTL, return the cached value immediately and start a background process to update the cache from the API.

The initial download should be intiated by a growthbook.loadFeatures() method call. This method may take optional parameters, such as timeout or skipCache, if it makes sense.

There should be an easy way for the user to wait until features finish loading. Depending on the language, this might be an event emitter, a Promise, a callback, or something similar. Use whatever method is standard for the language.

Server-Sent Events

The API response (/api/features/{clientKey}) may contain a response header:

x-sse-support: enabled

If set to "enabled", you are able to subscribe to the API for realtime changes to feature definitions by using the GrowthBook Proxy. This will let you update the cache immediately when a feature changes instead of waiting for the 60s TTL to expire.

The URL for subscribing to changes is {apiHost}/sub/{clientKey}.

SDKs should not attempt to subscribe to the /sub/ endpoint unless the header x-sse-support: enabled is present on the /api/features endpoint response.

An example implementation in JavaScript is below:

const channel = new EventSource(`${apiHost}/sub/${clientKey}`);
channel.addEventListener("features", (event) => {
const data = JSON.parse(event.data);
cache.set(clientKey, data.features);
})

Some important things to note:

  • The SDK should implement reconnect logic to support both client and server dropping the connection.
  • The response will be the same as when fetching from the /api/features/{clientKey} endpoint

Encrypted Features

The /api/features/{clientKey} endpoint can have encryption enabled (128-bit AES-CBC). When this is the case, the API response will look like this:

{
"features": {},
"encryptedFeatures": "abcdef123456.ghijklmnop789jksdkfaksfadfasdfkahsfa"
}

Before you can use this response, you will need to decrypt it. This requires the user to set Context.decryptionKey when creating the GrowthBook instance.

Type Hinting

Most languages have some sort of strong typing support, whether in the language itself or via annotations. This helps to reduce errors and is highly encouraged for SDKs.

If possible, use generics to type the return value. For example, if experiment.variations is type T[], then result.value should be type T. Or, if the fallback of getFeatureValue is type string, the return type should also be type string.

Handling Errors

The general rule is to be strict in development and lenient in production.

You can throw exceptions in development, but someone's production app should never crash because of a call to growthbook.evalFeature or growthbook.run.

For the below edge cases in production, just act as if the problematic property didn't exist and ignore errors:

  • experiment.weights is a different length from experiment.variations
  • experiment.weights adds up to something other than 1
  • experiment.coverage or feature.coverage is greater than 1 or less than 0
  • context.trackingCallback throws an error
  • URL querystring specifies an invalid variation index

For the below edge cases in production, the experiment should be disabled (everyone gets assigned variation 0):

  • experiment.coverage is less than 0
  • experiment.force specifies an invalid variation index
  • context.forcedVariations specifies an invalid variation index
  • experiment.hashAttribute is an empty string

Subscriptions

Sometimes it's useful to be able to "subscribe" to a GrowthBook instance and be alerted every time growthbook.run is called. This is different from the tracking callback since it also fires when a user is not included in an experiment.

growthbook.subscribe(function (experiment, result) {
// do something
});

It's best to only re-fire the callbacks for an experiment if the result has changed. That means either the inExperiment flag has changed or the variationId has changed.

If it makes sense for your language, this function should return an "unsubscriber". A simple callback that removes the subscription.

unsubscriber = growthbook.subscribe(...)
unsubscriber()

In addition to subscriptions you may also want to expose a growthbook.getAllResults method that returns a map of the latest results indexed by experiment key.

Memory Management

Subscriptions and tracking calls require storing references to many objects and functions. If it makes sense for your language, libraries should provide a growthbook.destroy method to remove all of these references and release their memory.

Tests

We strive to have 100% test coverage for all of our SDKs.

There is a language-agnostic test suite stored as a JSON file (https://github.com/growthbook/growthbook/blob/main/packages/sdk-js/test/cases.json) with more than 400 unit tests. This extensively tests all of the public methods mentioned above.

The cases.json file is an object. The keys are the function being tested, and the values are arrays of test cases. The test case arrays structure is different for each function and listed below:

  • evalCondition
    • name of the test case (string)
    • condition
    • attributes
    • expected return value (boolean)
    • definitions for Saved Groups referenced in the test case (object of keys: ID of list -> values: array of members)
  • hash
    • seed (string)
    • value to hash (string)
    • hash version to use (integer)
    • expected result (float)
  • getBucketRange
    • Name of the test case (string)
    • Arguments array ([numVariations, coverage, weights or null])
    • expected result
  • feature (evalFeature)
    • name of the test case (string)
    • context passed into GrowthBook constructor
    • name of the feature (string)
    • expected result
  • run
    • name of the test case (string)
    • context passed into GrowthBook constructor
    • experiment object
    • expected value
    • inExperiment (boolean)
    • hashUsed (boolean)
  • chooseVariation
    • name of the test case (string)
    • n (hash)
    • bucket ranges
    • expected result
  • getQueryStringOverride
    • name of the test case (string)
    • experiment key
    • url
    • numVariations
    • expected result
  • inNamespace
    • name of the test case (string)
    • id
    • namespace
    • expected result
  • getEqualWeights
    • numVariations
    • expected result (weights rounded to 8 decimal places)
  • decrypt
    • name of the test case (string)
    • encrypted text (string)
    • decryption key (string)
    • expected result (string or null if the decryption should fail)
  • stickyBucket
    • name of the test case (string)
    • context passed into GrowthBook constructor
    • array of preexisting sticky bucket assignment docs
    • name of the feature (string)
    • expected result
    • expected sticky bucket assignment docs
  • urlRedirect
    • name of the test case (string)
    • context passed into GrowthBook constructor
    • expected array of result objects

In addition to the above, you should write custom test cases for things like event subscriptions, tracking callbacks, getters/setters, etc. that are more language-specific.

Getting Help

Join our Slack community if you need help or want to chat. We're also happy to hop on a call and do some pair programming.

Attribution

Open a GitHub issue with a link to your project and we'll make sure we add it to our docs and give you proper credit for your hard work.

Changelog

  • v0.1 2022-05-23
    • Don't skip experiment rules that are forced
  • v0.2 2022-07-19
    • Add featureId to ExperimentResult object
  • v0.2.1 2022-08-01
    • Add test case for when an experiment's hashAttribute is null
  • v0.2.2 2022-09-08
    • Add test case for when an experiment's hashAttribute is an integer
  • v0.2.3 2022-12-06
    • Add test case for when an experiment's coverage is set to 0
  • v0.3.0 2023-01-18
    • New apiHost, clientKey, and decryptionKey Context properties
    • Built-in fetching and caching
    • Server Sent Events (SSE) support for realtime feature updates
  • v0.4.0 2023-02-24
    • Changed signature of hash method and added multiple hashing versions
    • New inRange, isIncludedInRollout, and isFilteredOut helper methods
    • New hashVersion, range, ranges, meta, filters, seed, name, tracks, and phase properties of FeatureRules
    • New hashVersion, ranges, meta, filters, seed, name, and phase properties of Experiments
    • New key, name, bucket, and passthrough fields in Experiment Results
    • New Filter, VariationMeta, and TrackData data structures
  • v0.4.1 2023-04-13
    • Added decrypt function and set of test cases
    • hash function now returns null instead of -1 when an invalid hashVersion is specified
    • Fixed broken feature test case (was using [0.99] instead of 0.99 for coverage)
  • v0.4.2 2023-04-30
    • Add test cases when targeting condition value is null
  • v0.5.0 2023-05-17
    • Add support for new version string comparison operators ($veq, $vne, $vgt, $vgte, $vlt, $vlte) and new paddedVersionString helper function
    • New isIn helper function for conditions, plus new evalCondition test cases for $in and $nin operators when attribute is an array
  • v0.5.1 2023-10-19
    • Add 2 new test cases for matching on a $groups array attribute
  • v0.5.2 2023-10-30
    • Add 3 new test cases for comparison operators to handle more edge cases
  • v0.5.3 2024-01-02
    • Experiment conditions are now evaluated within the experiment object
    • New fallbackAttribute, disableStickyBucketing, bucketVersion, minBucketVersion, properties of FeatureRules
    • New fallbackAttribute, disableStickyBucketing, bucketVersion, minBucketVersion, properties of Experiments
    • Add stickyBucketUsed to ExperimentResult object
  • v0.5.4 2024-02-23
    • New parentConditions property of FeatureRules
    • New parentConditions property of Experiments
  • v0.5.5 2024-04-09
    • Add test cases for URL Redirects
    • Add navigate method to Context
    • Add persistQueryString property to Experiment
    • Add and improve test cases for StickyBucket
  • v0.6.0 2024-04-30
    • Remove versionCompare test cases (these are now just included as part of evalCondition)
    • Tweak to isIncludedInRollout to handle an edge case when coverage is zero. Also added test case for this.
    • Add id property to feature rules (reserved for future use)
  • v0.6.1 2024-05-13
    • Update logic in evalCondition to allow for and/or/not/nor operators to appear at the same level as other conditions
    • Added test cases for multiple operators on the same level
  • v0.7.0 2024-06-25
    • New Operators $inGroup and $notInGroup to check Saved Groups by reference
    • Add argument to evalCondition for definition of Saved Groups
    • Add test cases for evalCondition, feature, and run using the new operators