feat: Avoid updating populated store with empty configs during init

[stephenreid]

Summary

This pull request introduces a safeguard to prevent the local configuration from being wiped when the network returns an empty set of specifications while a data store is in use. This ensures that the SDK continues to use the last known good configuration if the network sync fetches an empty response.

Changes

lib/spec_store.rb

• Added is_empty_specs?(specs_json): A private helper method to determine if a given specifications object contains no feature gates, dynamic configs, or layer configs.
• Updated process_specs(specs_string, from_adapter: false):
  • Added a check to see if the incoming specifications are empty.
  • If the specs are empty, the network is the source (from_adapter is false), and a data store is configured, the update is skipped (returns nil early).

test/data_adapter_test.rb

• Added test_datastore_preserves_data_when_network_returns_empty_specs:
  • A new test case that simulates a network response returning empty specifications.
  • Verifies that both the SpecStore's internal state and the configured DataStore preserve the original configuration instead of being overwritten by the empty response.

Test Plan

• Run test_datastore_preserves_data_when_network_returns_empty_specs to verify data preservation.
• Ensure existing data adapter tests still pass.

Impact

This change improves the resilience of the SDK when using a persistent data store. It prevents a "blank slate" scenario where all feature gates and configurations would temporarily evaluate to their default values if the network sync happens to fetch an empty response.

[stephenreid]

@lfoster-statsig For consideration for those of us who have implemented Redis (or other persistence) with an iDataStore interface to share across multi-instance envs

[lfoster-statsig]

@lfoster-statsig For consideration for those of us who have implemented Redis (or other persistence) with an iDataStore interface to share across multi-instance envs

Honestly do think this is a beneficial idea, will get more eyes internally

[stephenreid]

The test failure appears to be a change in how the sdk key is passed in the test/ci environment between 2.5.0 and 3.2.0

ruby-sdk/test/test_country_lookup.rb

Line 43 in e78d90e

secret = ENV['test_api_key']

[stephenreid]

The bootstrap and adapter choice already choose this