Feature Flags

When a new feature is added to Ember they will be written in such a way that the feature can be conditionally included in the generated build output and enabled (or completely removed) based on whether a particular flag is present. This allows newly developed features to be selectively released when they are considered ready for production use.

When a new feature is flagged it is only available in canary builds (if enabled at runtime). When it is time for the next beta cycle to be started (generally 6-12 week cycles) each feature will be evaluated and those features that are ready will be enabled in the next beta (and subsequently automatically enabled in all future canary builds).

If a given feature is deemed unstable it will be disabled in the next beta point release, and not be included in the next stable release. It may still be included in the next beta cycle if the issues/concerns have been resolved.

Once the beta cycle has completed the final release will include any features that were enabled during that cycle. At this point the feature flags will be removed from the canary and future beta branches, and the feature flag will no longer be used.

The flag status in the generated build output is controlled by the features.json file in the root of the project. This file lists all features and their current status.

A feature can have one of a few different statuses:

• true - The feature is enabled: the code behind the flag is always enabled in the generated build.
• false - The feature is disabled: the code behind the flag is not present in the generated build at all.
• null - The feature is present in the build output, but must be enabled at runtime (it is still behind feature flags).

The process of removing the feature flags from the resulting build output is handled by defeatureify.

When a new feature is added to the canary channel (aka master branch), an entry is added to FEATURES.md explaining what the feature does (and linking the originating pull request). This listing is kept current, and reflects what is available in each branch (stable,beta, and master).

The only time a feature can be enabled at runtime is if the features.json for that build contains null (technically, anything other than true or false will do, but null is the chosen value).

A global EmberENV object will be used to initialize the Ember.ENV object, and any feature flags that are enabled/disabled under EmberENV.FEATURES will be migrated to Ember.FEATURES; those features will be enabled based on the flag value. Ember only reads the EmberENV value upon initial load so setting this value after Ember has been loaded will have no affect.

Example:

EmberENV = {FEATURES: {'link-to': true}};

Additionally you can define EmberENV.ENABLE_ALL_FEATURES to force all features to be enabled.