Module: Shipeasy

Defined in:
lib/shipeasy/admin.rb,
lib/shipeasy-sdk.rb,
lib/shipeasy/admin.rb,
lib/shipeasy/client.rb,
lib/shipeasy/config.rb,
lib/shipeasy/engine.rb,
lib/shipeasy/logging.rb,
lib/shipeasy/sdk/env.rb,
lib/shipeasy/sdk/see.rb,
lib/shipeasy/sdk/eval.rb,
lib/shipeasy/sdk/skill.rb,
lib/shipeasy/sdk/anon_id.rb,
lib/shipeasy/sdk/murmur3.rb,
lib/shipeasy/sdk/railtie.rb,
lib/shipeasy/sdk/version.rb,
lib/shipeasy/i18n/railtie.rb,
lib/shipeasy/sdk/telemetry.rb,
lib/shipeasy/sdk/openfeature.rb,
lib/shipeasy/sdk/sticky_store.rb,
lib/shipeasy/i18n/view_helpers.rb,
lib/shipeasy/i18n/label_fetcher.rb,
lib/shipeasy/sdk/internal_report.rb,
lib/shipeasy/sdk/rack_middleware.rb,
lib/generators/shipeasy/install/install_generator.rb

Overview

Leveled logger shared by the whole gem.

Every diagnostic the SDK emits from a caught error goes through here, so a single log_level config option (default :warn) controls the SDK's stderr output. The contract for the SDK's public RUNTIME methods (get_flag, get_config, universe(...).assign, get_killswitch, track, see, …) is that they NEVER raise into product code — so logging itself is best-effort too: a broken/throwing $stderr can never take down a flag read.

Level ordering (a message at level L is emitted iff the configured level is at least as verbose as L):

silent < error < warn < info < debug

:warn (the default) therefore emits error + warn and suppresses the informational info / debug chatter.

Defined Under Namespace

Modules: Admin, Generators, I18n, Logging, OpenFeature, SDK Classes: Client, Configuration, Engine, Error

Class Method Summary collapse

Class Method Details

.apply_overrides(engine, flags, configs, experiments) ⇒ Object

Apply the configure_for_* override args onto an engine.



323
324
325
326
327
328
329
330
 
# File 'lib/shipeasy/config.rb', line 323

def apply_overrides(engine, flags, configs, experiments)
  (flags || {}).each { |name, value| engine.override_flag(name, value) }
  (configs || {}).each { |name, value| engine.override_config(name, value) }
  (experiments || {}).each do |name, spec|
    group, params = spec   # spec is [group, params]
    engine.override_experiment(name, group, params)
  end
end

.attributes_transform ⇒ Object

The resolved attributes transform (callable). Default = identity, so a user object that is already the attribute hash is used verbatim.



142
143
144
145
146
147
148
149
150
151
 
# File 'lib/shipeasy/config.rb', line 142

def attributes_transform
  transform = config.attributes
  if transform.nil?
    ->(user) { user }
  elsif transform.respond_to?(:call)
    transform
  else
    raise Error, "Shipeasy.configure { |c| c.attributes = … } must be a callable (e.g. a lambda)"
  end
end

.bootstrap_script_tag(user, anon_id: nil, i18n_profile: "en:prod", base_url: nil) ⇒ Object 



291
292
293
294
295
 
# File 'lib/shipeasy/config.rb', line 291

def bootstrap_script_tag(user, anon_id: nil, i18n_profile: "en:prod", base_url: nil)
  require_engine("bootstrap_script_tag").bootstrap_script_tag(
    user, anon_id: anon_id, i18n_profile: i18n_profile, base_url: base_url
  )
end

.clear_overrides ⇒ Object

Drop EVERY override — including the seed from configure_for_testing (test mode has no blob beneath); under configure_for_offline it reverts to the snapshot.



272
273
274
275
 
# File 'lib/shipeasy/config.rb', line 272

def clear_overrides
  require_engine("clear_overrides").clear_overrides
  nil
end

.config ⇒ Object 



115
116
117
 
# File 'lib/shipeasy/config.rb', line 115

def config
  @config ||= Configuration.new
end

.configure {|config| ... } ⇒ Object

Configure the gem once at boot. In addition to populating the shared Configuration, this builds and registers the ONE global Shipeasy::Engine (first-config-wins) from the api_key/base_url and kicks off its one-shot fetch (fire-and-forget) so Shipeasy::Client.new(user).get_flag(...) resolves against real rules with no explicit init call.

Shipeasy.configure do |c|
c.api_key    = ENV["SHIPEASY_SERVER_KEY"]
c.attributes = ->(u) { { "user_id" => u.id, "plan" => u.plan } }
end

Shipeasy::Client.new(current_user).get_flag("new_checkout")

Long-running servers that also want the background poll can call Shipeasy.engine.init after configure.

Yields:



134
135
136
137
138
 
# File 'lib/shipeasy/config.rb', line 134

def configure
  yield config
  register_engine!(config) if config.api_key
  config
end

.configure_for_offline(snapshot: nil, path: nil, flags: nil, configs: nil, experiments: nil, attributes: nil) ⇒ Object

Configure Shipeasy OFFLINE — evaluate the REAL rules from an in-memory snapshot or a JSON file, with no network. Provide exactly one source:

snapshot: { "flags" => <body of /sdk/flags>, "experiments" => <body of /sdk/experiments> }
path:     "snapshot.json"   (a JSON file of the same shape)

Optional flags/configs/experiments overrides layer on top (same shapes as configure_for_testing). Replaces any previously-configured engine.



235
236
237
238
239
240
241
242
243
244
245
246
247
 
# File 'lib/shipeasy/config.rb', line 235

def configure_for_offline(snapshot: nil, path: nil, flags: nil, configs: nil, experiments: nil, attributes: nil)
  engine =
    if path
      Engine.from_file(path)
    elsif snapshot
      s = snapshot.transform_keys(&:to_s)
      Engine.from_snapshot(flags: s["flags"], experiments: s["experiments"])
    else
      raise Error, "Shipeasy.configure_for_offline requires snapshot: or path:"
    end
  apply_overrides(engine, flags, configs, experiments)
  install_global_engine(engine, attributes)
end

.configure_for_testing(flags: nil, configs: nil, experiments: nil, attributes: nil) ⇒ Object

Configure Shipeasy in TEST MODE — no api key, zero network, ever. Seed the values your code under test should see via the override args, then read through the ordinary Shipeasy::Client.new(user):

Shipeasy.configure_for_testing(flags: { "new_checkout" => true })
Shipeasy::Client.new({ "user_id" => "u_1" }).get_flag("new_checkout") # => true

flags:       { name => bool }              forced get_flag results
configs:     { name => value }             forced get_config results
experiments: { name => [group, params] }   forced enrolments
attributes:  same transform as configure (default identity)


221
222
223
224
225
 
# File 'lib/shipeasy/config.rb', line 221

def configure_for_testing(flags: nil, configs: nil, experiments: nil, attributes: nil)
  engine = Engine.for_testing
  apply_overrides(engine, flags, configs, experiments)
  install_global_engine(engine, attributes)
end

.control_flow_exception(err) ⇒ Object 



308
309
310
 
# File 'lib/shipeasy/config.rb', line 308

def control_flow_exception(err)
  Shipeasy::SDK.control_flow_exception(err)
end

.engine ⇒ Object

The single global engine registered by configure, or nil if configure has not run (or ran without an api_key). Shipeasy::Client reads this.



155
156
157
158
159
160
161
162
163
164
 
# File 'lib/shipeasy/config.rb', line 155

def engine
  pid = Process.pid
  if @engine && @engine_pid != pid
    # Post-fork: the parent's poll thread didn't survive. Rebuild lazily
    # from the stored config in this child process.
    @engine = nil
    register_engine!(config) if config.api_key
  end
  @engine
end

.flags ⇒ Object

Lazy, fork-safe singleton Engine. The first call from each process spawns a fresh client + poll thread — including post-fork workers under Puma's preload_app!. Callers can Shipeasy.flags.get_flag(...) straight from a controller without holding a constant or worrying about before_worker_boot hooks.

Initializers stay minimal:

# config/initializers/shipeasy.rb
Shipeasy.configure { |c| c.api_key = ENV["SHIPEASY_SERVER_KEY"] }

The first request that touches Shipeasy.flags.* triggers init(). For serverless / Lambda where you want a single fetch with no thread, build the engine explicitly: Shipeasy::Engine.new(...).init_once.

NOTE: this remains a separate, polling engine from the one configure() registers (Shipeasy.engine). New code should prefer the Shipeasy.configure + Shipeasy::Client.new(user) front door; Shipeasy.flags is retained for the legacy Shipeasy.flags.get_flag(name, user) style.



380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
 
# File 'lib/shipeasy/config.rb', line 380

def flags
  pid = Process.pid
  if @flags && @flags_pid != pid
    # Post-fork: parent's poll thread didn't survive. Don't destroy
    # @flags (its mutex/state is invalid in this child anyway); just
    # rebuild from scratch.
    @flags = nil
  end
  @flags ||= begin
    @flags_pid = pid
    client = Engine.new(
      api_key:  config.api_key,
      base_url: config.base_url,
    )
    client.init
    client
  end
end

.i18n_script_tag(client_key, profile: "en:prod", base_url: nil) ⇒ Object

SSR tag helpers — delegate to the configured global engine, so you never touch it. i18n_script_tag carries the PUBLIC client key (not the server key); bootstrap_script_tag embeds no key.



287
288
289
 
# File 'lib/shipeasy/config.rb', line 287

def i18n_script_tag(client_key, profile: "en:prod", base_url: nil)
  require_engine("i18n_script_tag").i18n_script_tag(client_key, profile: profile, base_url: base_url)
end

.install_global_engine(engine, attributes) ⇒ Object

Replace the registered global engine + attributes transform (used by the configure_for_* siblings — unlike configure, they replace so a test suite can reconfigure between cases). Returns the engine.



315
316
317
318
319
320
 
# File 'lib/shipeasy/config.rb', line 315

def install_global_engine(engine, attributes)
  config.attributes = attributes
  @engine = engine
  @engine_pid = Process.pid
  engine
end

.on_change(callable = nil, &block) ⇒ Object

Register a poll listener fired after a background poll fetches NEW data (HTTP 200, not 304). Requires configure(poll: true). Returns an unsubscribe proc. Accepts a block or any callable.



280
281
282
 
# File 'lib/shipeasy/config.rb', line 280

def on_change(callable = nil, &block)
  require_engine("on_change").on_change(callable, &block)
end

.override_config(name, value) ⇒ Object 



259
260
261
262
 
# File 'lib/shipeasy/config.rb', line 259

def override_config(name, value)
  require_engine("override_config").override_config(name, value)
  nil
end

.override_experiment(name, group, params) ⇒ Object 



264
265
266
267
 
# File 'lib/shipeasy/config.rb', line 264

def override_experiment(name, group, params)
  require_engine("override_experiment").override_experiment(name, group, params)
  nil
end

.override_flag(name, value) ⇒ Object

On-the-spot overrides layered on top of whatever configure_for_testing / configure_for_offline (or a live configure) set up — they win over the blob until clear_overrides. Require a prior configure* call.



254
255
256
257
 
# File 'lib/shipeasy/config.rb', line 254

def override_flag(name, value)
  require_engine("override_flag").override_flag(name, value)
  nil
end

.register_engine!(cfg) ⇒ Object

Build + register the one global engine (first-config-wins). Kicks off the configured fetch lifecycle (one-shot by default; the background poll when c.poll = true) fire-and-forget. Idempotent within a process.



169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
 
# File 'lib/shipeasy/config.rb', line 169

def register_engine!(cfg)
  return @engine if @engine && @engine_pid == Process.pid
  @engine_pid = Process.pid
  engine = Engine.new(
    api_key:            cfg.api_key,
    base_url:           cfg.base_url,
    env:                cfg.env,
    is_network_enabled: cfg.is_network_enabled,
    disable_telemetry:  cfg.disable_telemetry,
    telemetry_url:      cfg.telemetry_url,
    private_attributes: cfg.private_attributes,
    sticky_store:       cfg.sticky_store,
    log_level:          cfg.log_level,
    disable_internal_error_reporting: cfg.disable_internal_error_reporting,
  )
  @engine = engine
  # Capture +engine+ in the closure (not the @engine ivar, which a concurrent
  # reset/reconfigure could nil out before the thread runs).
  if cfg.poll
    @boot_thread = Thread.new do
      engine.init   # initial fetch + background poll thread
    rescue => e
      Shipeasy::Logging.error "[shipeasy] configure(poll) background poll failed: #{e.message}"
    end
  elsif cfg.init
    @boot_thread = Thread.new do
      engine.init_once
    rescue => e
      Shipeasy::Logging.error "[shipeasy] configure() one-shot fetch failed: #{e.message}"
    end
  end
  engine
end

.require_engine(fn_name) ⇒ Object

The global engine, or raise a helpful error naming the package-level fn the caller used before any configure*.

Raises:



334
335
336
337
338
339
340
341
 
# File 'lib/shipeasy/config.rb', line 334

def require_engine(fn_name)
  e = engine
  return e unless e.nil?

  raise Error, "Shipeasy.#{fn_name} called before Shipeasy.configure " \
               "{ |c| c.api_key = … } (or configure_for_testing / " \
               "configure_for_offline). Call one once at app boot."
end

.reset_config! ⇒ Object

Reset the config back to defaults — primarily for tests.



344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
 
# File 'lib/shipeasy/config.rb', line 344

def reset_config!
  # Reap the fire-and-forget boot thread first: it holds its own engine
  # reference, so left alive it can call init/init_once during the NEXT
  # test example and trip any_instance expectations armed there.
  if (t = @boot_thread)
    t.join(1) || t.kill
    @boot_thread = nil
  end
  @config = nil
  @flags_pid = nil
  @flags&.destroy
  @flags = nil
  @engine&.destroy
  @engine = nil
  @engine_pid = nil
end

.see(problem) ⇒ Object

see() structured error reporting — package-level, dispatched through the last-constructed default client (the engine configure built). Never raises into caller code; a call before any client exists warns and no-ops.



300
301
302
 
# File 'lib/shipeasy/config.rb', line 300

def see(problem)
  Shipeasy::SDK.see(problem)
end

.see_violation(name) ⇒ Object 



304
305
306
 
# File 'lib/shipeasy/config.rb', line 304

def see_violation(name)
  Shipeasy::SDK.see_violation(name)
end