Feature flag client

This client makes it easy to work with feature flags and dark features. By using it, exposure events will be fired automatically allowing analysis of important metrics out of the box.

Usage

Bootstrap

import FrontendFeatureFlagClient from '@atlaskit/feature-flag-client';

const client = new FrontendFeatureFlagClient({
  analyticsHandler: myAnalyticsHandler,
  flags: {
    'my.experiment': {
      value: 'experiment',
      explanation: {
        kind: 'RULE_MATCH',
        ruleId: '111-bbbbb-ccc',
      },
    },
    'my.boolean.flag': {
      value: false,
    },
    'my.json.flag': {
      value: {
        nav: 'blue',
        footer: 'black',
      },
      explanation: {
        kind: 'RULE_MATCH',
        ruleId: '111-bbbbb-ccc',
      },
    },
    'my.detailed.boolean.flag': {
      value: false,
      explanation: {
        kind: 'RULE_MATCH',
        ruleId: '111-bbbbb-ccc',
      },
    },
  },
});

Retrieving values

// flag set, returns real value
client.getBooleanValue('my.detailed.boolean.flag', { default: true }); // > false

// flag set, returns real value
client.getVariantValue('my.experiment', {
  default: 'control',
  oneOf: ['control', 'experiment'],
}); // > experiment

// flag unset, returns default value
client.getBooleanValue('my.unlisted.boolean.flag', { default: false }); // > false

// flag value doesn't match expected, returns default
client.getVariantValue('my.experiment', {
  default: 'control',
  oneOf: ['control', 'variant-a'],
}); // > control

// flag set, returns real value
client.getJSONFlag('my.json.flag'); // > { nav: 'blue', footer: 'black' }

// flag set, returns real value
client.getRawValue('my.delimited.flag', {
  default: 'dashboard,issue-view,classic-board',
}); // > dashboard,issue-view,classic-board,next-gen-board

NOTE: For performance reasons, the client will only run validation checks once and then cache this result for subsequent calls. It is important that you always use the same method and oneOf values whenever retrieving the value for your flag.

Setting flags asynchronously

If you load your flags after the app bootstrap, you set the to the client through the 'setFlags' method.

client.setFlags({
  'my.async.boolean.flag': {
    value: false,
    explanation: {
      kind: 'RULE_MATCH',
      ruleId: '333-bbbbb-ccc',
    },
  },
});

Exposure events

Exposure events are great for tracking which users and sites saw your flags and in what state. These exposure events can be used for:

1. Rollouts,
2. Experiments, and
3. Incident investigations.

To send any exposure events, you need to provide an object that looks like src/types.ts#AnalyticsHandler (or internally to Atlassian this is the same as the AnalyticsWebClient).

This can be set on initialisation of the client, or later with the client.setAnalyticsHandler(analyticsHandler); fucntion.

How to send exposure events for all feature flags

To turn on Automatic Exposures, you can either:

1. Set isAutomaticExposuresEnabled to true in the constructor args, or
2. Call client.setIsAutomaticExposuresEnabled(true).

When this mode is enabled, exposure events will be handled in the following circumstances: sent for all simple flags (ie. flags without an explanation), and for all evalulated (ie. flags with evalauation details) with the shouldTrackExposureEvent option set to false.

• Simple Flags: Automatic exposure events will always be fired with the flag key and value
• Evaluated Flags with shouldTrackExposureEvent: false: Automatic exposure events will fire with explanation details
• Evaluated Flags with shouldTrackExposureEvent: true: Automatic exposure events will not fire, but the regular ones will.

The exposure events that are fired in this fashion will be tagged with an additional tag on the event tags: ['autoExposure'] to make it easier to differientiate between those exposures fired manually and those fired automatically.

How to avoid firing the exposure event

You can skip the exposure event by setting 'shouldTrackExposureEvent' to 'false'. Note: this will not disable the automatic exposure event from firing when the Automode is enabled.

client.getBooleanValue('my.detailed.boolean.flag', {
  default: true,
  shouldTrackExposureEvent: false,
});

How to fire the exposure event manually

client.trackExposure('my.detailed.boolean.flag', {
  value: true,
  explanation: {
    kind: 'RULE_MATCH',
    ruleId: 'aaaa-vbbbb-ccccc',
  },
});

How to include custom attributes in the exposure event

You can send extra attributes by including an object within the exposureData attribute. This object accepts strings, numbers or booleans. Reserved attributes can't be used, like flagKey, reason, ruleId or value.

client.getBooleanValue('my.detailed.boolean.flag', {
  default: true,
  shouldTrackExposureEvent: false,
  exposureData: {
    myAttribute1: 'whatever',
    myAttribute2: 2,
    myAttribute3: true,
  },
});

If you are firing the exposure event manually, the 3rd argument (optional) includes the custom attributes. This object accepts strings, numbers or booleans. Reserved attributes can't be used, like flagKey, reason, ruleId or value.

client.trackExposure(
  'my.detailed.boolean.flag',
  {
    value: true,
    explanation: {
      kind: 'RULE_MATCH',
      ruleId: 'aaaa-vbbbb-ccccc',
    },
  },
  {
    myAttribute1: 'whatever',
    myAttribute2: 2,
    myAttribute3: true,
  },
);

Viewing flag stats

The getFlagStats method can be called to collate some information about client usage. This method returns an object that maps each flag key to an object describing how it has been accessed.

The only statistic that is currently returned for each flag is an evaluationCount, which indicates how many times the flag has been evaluated through any of the retrieval methods.

client.getVariantValue('my.experiment', {
  default: 'control',
  oneOf: ['control', 'experiment'],
});

client.getFlagStats(); // { 'my.experiment' : { evaluationCount: 1 }}

If you reset some flags using setFlags or clear, the stats collected for these flags will also be reset.