Module: Lucerna::Gates::Evaluate

Defined in:
lib/lucerna/gates/evaluate.rb

Overview

The Gates evaluation engine — pure functions over a compiled runtime, no I/O. Ported line-for-line from @lucerna/gates-core (sdks/core/src/evaluate.ts) onto the frozen bucketing contract in Hashing; the golden vectors in sdks/core/src/vectors.json pin parity.

The runtime document keeps its wire shape (string keys, camelCase); decisions come out symbol-keyed with reason strings identical to the wire codes.

Fail-safe rule: a gate carrying any rule element this engine version doesn't recognize (unknown condition operator, unknown override kind) is answered whole with "unsupported_rule" — flag off, experiment unassigned. An engine never half-evaluates a document it doesn't fully understand.

Constant Summary collapse

OPERATORS = 
%w[is is_not contains gt lt is_set].freeze
AUDIENCE_OPERATORS =

Flag conditions may additionally reference a saved audience: the compiled condition carries the audience id in "value" ("trait" is the "audience" sentinel) and matches when the identity's traits satisfy the audience's rule set.

%w[in_audience not_in_audience].freeze
OVERRIDE_BY = 
%w[user_id domain].freeze
OVERRIDE_SERVE = 
%w[on off].freeze
OVERRIDE_TRAIT_PREFIX =

Overrides can also pin by any trait the workspace enabled as an override dimension: "by" carries "trait:" and matches when the identity trait equals "who" exactly (same value-based compare as the "is" operator; "domain" keeps its lowercase compare).

"trait:"

Class Method Summary collapse

Class Method Details

.all(runtime, user_id: nil, traits: nil) ⇒ Object

One identity's decisions over the entire runtime.



252
253
254
255
256
257
258
259
260
261
262
 
# File 'lib/lucerna/gates/evaluate.rb', line 252

def all(runtime, user_id: nil, traits: nil)
  {
    kills: runtime["kills"].dup,
    flags: runtime["flags"].to_h do |key, value|
      [key, flag(value, user_id: user_id, traits: traits, audiences: runtime["audiences"])]
    end,
    experiments: runtime["experiments"].to_h do |key, value|
      [key, experiment(value, runtime["audiences"], user_id: user_id, traits: traits)]
    end,
  }
end

.assigned(variant, iteration) ⇒ Object 



270
271
272
273
274
275
276
 
# File 'lib/lucerna/gates/evaluate.rb', line 270

def assigned(variant, iteration)
  {
    variant: { id: variant["id"], name: variant["name"], is_control: variant["isControl"] },
    iteration: iteration,
    reason: "assigned",
  }
end

.compare_values(a, b) ⇒ Object

Ordering for gt/lt: numeric when both sides parse as numbers, otherwise lexicographic by UTF-16 code units (ISO dates sort correctly). Code-unit comparison — not locale collation — is part of the frozen contract.



282
283
284
285
286
287
288
 
# File 'lib/lucerna/gates/evaluate.rb', line 282

def compare_values(a, b)
  numeric_a = js_number(a)
  numeric_b = js_number(b)
  return numeric_a <=> numeric_b if numeric_a && numeric_b

  a.encode(Encoding::UTF_16BE).b <=> b.encode(Encoding::UTF_16BE).b
end

.experiment(experiment, audiences, user_id: nil, traits: nil, trace: nil) ⇒ Object

Experiment assignment, sticky per (salt, user): eligibility gates first (running, environment, audience), then the holdout and traffic dice, then an allocation-weighted walk over the variants. The salt is per iteration, so restarting an iteration reshuffles every bucket by design.



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
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
 
# File 'lib/lucerna/gates/evaluate.rb', line 169

def experiment(experiment, audiences, user_id: nil, traits: nil, trace: nil)
  iteration = experiment["iteration"]

  unless experiment["running"]
    note(trace, "status", "experiment is not running")
    return no_variant(iteration, "not_running")
  end
  note(trace, "status", "experiment is running")
  unless experiment["enabled"]
    note(trace, "environment", "experiment is off in this environment")
    return no_variant(iteration, "environment_off")
  end
  note(trace, "environment", "experiment is on in this environment")

  traits ||= {}
  audience_id = experiment["audienceId"]
  if audience_id.nil?
    note(trace, "audience", "targets everyone")
  else
    audience = audiences[audience_id]
    unless audience
      note(trace, "audience", %(audience "#{audience_id}" is not in the runtime))
      return no_variant(iteration, "audience_not_found")
    end
    bad_operator = unsupported_operator(audience["conditions"])
    if bad_operator
      note(
        trace,
        "unsupported",
        %(audience condition operator "#{bad_operator}" is not supported by this engine),
      )
      return no_variant(iteration, "unsupported_rule")
    end
    in_audience = matches_audience?(traits, audience)
    note(
      trace,
      "audience",
      %(audience "#{audience_id}" (#{audience["match"]}): #{in_audience ? "matched" : "no match"}),
    )
    return no_variant(iteration, "not_in_audience") unless in_audience
  end

  if user_id.nil? || user_id.empty?
    note(trace, "identity", "no userId — assignments have nothing to be sticky by")
    return no_variant(iteration, "missing_user_id")
  end

  holdout = experiment["holdout"]
  if holdout.positive?
    roll = Hashing.bucket("#{experiment["salt"]}:holdout", user_id)
    held = roll < holdout * 100
    note(trace, "holdout", "bucket #{roll} vs holdout #{holdout * 100}: #{held ? "held out" : "in"}")
    return no_variant(iteration, "holdout") if held
  end
  traffic = experiment["traffic"]
  if traffic < 100
    roll = Hashing.bucket("#{experiment["salt"]}:traffic", user_id)
    out = roll >= traffic * 100
    note(trace, "traffic", "bucket #{roll} vs traffic #{traffic * 100}: #{out ? "out" : "in"}")
    return no_variant(iteration, "not_in_traffic") if out
  end

  # Allocations sum to 100 (enforced on write); the last variant
  # absorbs any rounding gap so the walk always lands.
  roll = Hashing.bucket("#{experiment["salt"]}:variant", user_id)
  cumulative = 0
  experiment["variants"].each do |variant|
    cumulative += variant["allocation"] * 100
    if roll < cumulative
      note(trace, "variant", %(bucket #{roll} lands in "#{variant["name"]}"))
      return assigned(variant, iteration)
    end
  end
  last = experiment["variants"].last
  unless last
    note(trace, "variant", "experiment has no variants")
    return no_variant(iteration, "not_running")
  end
  note(trace, "variant", %(bucket #{roll} absorbed by last variant "#{last["name"]}"))
  assigned(last, iteration)
end

.flag(flag, user_id: nil, traits: nil, trace: nil, audiences: {}) ⇒ Object

Flag precedence, first hit wins:

1. environment master switch off -> off
2. overrides, in stored order (by user id or email domain)
3. conditions — all must match (a flag's rule is a conjunction)
4. percentage rollout, sticky by user id via the stored salt

A rollout between 0 and 100 with no user id serves off — there is nothing to be sticky by.

audiences resolves the flag's audience conditions (in_audience / not_in_audience); a referenced id that is absent evaluates the condition to no-match — never to a wider serve.



49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
 
# File 'lib/lucerna/gates/evaluate.rb', line 49

def flag(flag, user_id: nil, traits: nil, trace: nil, audiences: {})
  unless flag["enabled"]
    note(trace, "enabled", "flag is off in this environment")
    return { on: false, reason: "flag_off" }
  end
  note(trace, "enabled", "flag is on in this environment")

  bad_override = flag["overrides"].find do |override|
    !supported_override_by?(override["by"]) || !OVERRIDE_SERVE.include?(override["serve"])
  end
  bad_operator = unsupported_flag_operator(flag["conditions"])
  if bad_override || bad_operator
    note(
      trace,
      "unsupported",
      if bad_operator
        %(condition operator "#{bad_operator}" is not supported by this engine)
      else
        %(override kind "#{bad_override["by"]}/#{bad_override["serve"]}" is not supported by this engine)
      end,
    )
    return { on: false, reason: "unsupported_rule" }
  end

  traits ||= {}
  domain = identity_domain(traits)&.downcase
  flag["overrides"].each do |override|
    matched =
      case override["by"]
      when "user_id"
        !user_id.nil? && override["who"] == user_id
      when "domain"
        !domain.nil? && override["who"].downcase == domain
      else
        raw = traits[override["by"].delete_prefix(OVERRIDE_TRAIT_PREFIX)]
        !raw.nil? && raw == override["who"]
      end
    note(
      trace,
      "override",
      %(#{override["by"]} "#{override["who"]}" serves #{override["serve"]}: #{matched ? "matched" : "no match"}),
    )
    return { on: override["serve"] == "on", reason: "override" } if matched
  end

  unless flag["conditions"].empty?
    all = true
    flag["conditions"].each do |condition|
      if AUDIENCE_OPERATORS.include?(condition["operator"])
        audience = audiences[condition["value"]]
        if audience
          bad_audience_operator = unsupported_operator(audience["conditions"])
          if bad_audience_operator
            note(
              trace,
              "unsupported",
              %(audience condition operator "#{bad_audience_operator}" is not supported by this engine),
            )
            return { on: false, reason: "unsupported_rule" }
          end
        end
        # A dangling audience id never matches — for both operators the
        # condition fails closed rather than widening the serve.
        in_audience = !audience.nil? && matches_audience?(traits, audience)
        matched =
          !audience.nil? &&
          (condition["operator"] == "in_audience" ? in_audience : !in_audience)
        note(
          trace,
          "condition",
          if audience
            %(audience "#{condition["value"]}" #{condition["operator"] == "in_audience" ? "includes" : "excludes"} user: #{matched ? "matched" : "no match"})
          else
            %(audience "#{condition["value"]}" is not in the runtime: no match)
          end,
        )
      else
        matched = matches_condition?(traits, condition)
        note(
          trace,
          "condition",
          %(#{condition["trait"]} #{condition["operator"]} "#{condition["value"]}": #{matched ? "matched" : "no match"}),
        )
      end
      unless matched
        all = false
        break
      end
    end
    return { on: true, reason: "conditions" } if all
  end

  percentage = flag["rollout"]["percentage"]
  if percentage >= 100
    note(trace, "rollout", "rollout at 100% — everyone is in")
    return { on: true, reason: "rollout" }
  end
  if percentage <= 0
    note(trace, "rollout", "rollout at 0% — no one is in")
    return { on: false, reason: "rollout" }
  end
  if user_id.nil? || user_id.empty?
    note(trace, "rollout", "partial rollout with no userId — nothing to be sticky by")
    return { on: false, reason: "missing_user_id" }
  end
  roll = Hashing.bucket(flag["rollout"]["salt"], user_id)
  on = roll < percentage * 100
  note(
    trace,
    "rollout",
    "bucket #{roll} #{on ? "<" : ">="} threshold #{percentage * 100} (basis points of 10000)",
  )
  { on: on, reason: "rollout" }
end

.identity_domain(traits) ⇒ Object

The domain an override's by: "domain" matches: an explicit domain trait, falling back to the part of the email trait after the last "@".



344
345
346
347
348
349
350
351
352
353
 
# File 'lib/lucerna/gates/evaluate.rb', line 344

def identity_domain(traits)
  domain = traits["domain"]
  return domain if domain && !domain.empty?

  email = traits["email"]
  return nil if email.nil? || email.empty?

  at = email.rindex("@")
  at ? email[(at + 1)..] : nil
end

.js_number(value) ⇒ Object

Mirrors JavaScript Number(string) for the realistic trait space: whitespace-trimmed, empty string is 0, otherwise Float parsing.



292
293
294
295
296
297
298
299
 
# File 'lib/lucerna/gates/evaluate.rb', line 292

def js_number(value)
  stripped = value.strip
  return 0.0 if stripped.empty?

  Float(stripped)
rescue ArgumentError
  nil
end

.matches_audience?(traits, audience) ⇒ Boolean

Returns:

  • (Boolean) 


355
356
357
358
359
360
361
362
363
364
 
# File 'lib/lucerna/gates/evaluate.rb', line 355

def matches_audience?(traits, audience)
  conditions = audience["conditions"]
  return true if conditions.empty?

  if audience["match"] == "all"
    conditions.all? { |condition| matches_condition?(traits, condition) }
  else
    conditions.any? { |condition| matches_condition?(traits, condition) }
  end
end

.matches_condition?(traits, condition) ⇒ Boolean

Returns:

  • (Boolean) 


301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
 
# File 'lib/lucerna/gates/evaluate.rb', line 301

def matches_condition?(traits, condition)
  raw = traits[condition["trait"]]
  value = condition["value"]
  case condition["operator"]
  when "is_set"
    !raw.nil? && raw != ""
  when "is"
    raw == value
  when "is_not"
    raw != value
  when "contains"
    (raw || "").downcase.include?(value.downcase)
  when "gt"
    !raw.nil? && compare_values(raw, value).positive?
  when "lt"
    !raw.nil? && compare_values(raw, value).negative?
  else
    false # unreachable — callers gate on unsupported_operator
  end
end

.no_variant(iteration, reason) ⇒ Object

-- internals ------------------------------------------------------



266
267
268
 
# File 'lib/lucerna/gates/evaluate.rb', line 266

def no_variant(iteration, reason)
  { variant: nil, iteration: iteration, reason: reason }
end

.note(trace, step, detail) ⇒ Object 



366
367
368
 
# File 'lib/lucerna/gates/evaluate.rb', line 366

def note(trace, step, detail)
  trace << { step: step, detail: detail } if trace
end

.supported_override_by?(by) ⇒ Boolean

Returns:

  • (Boolean) 


327
328
329
330
331
 
# File 'lib/lucerna/gates/evaluate.rb', line 327

def supported_override_by?(by)
  return true if OVERRIDE_BY.include?(by)

  by.start_with?(OVERRIDE_TRAIT_PREFIX) && by.length > OVERRIDE_TRAIT_PREFIX.length
end

.unsupported_flag_operator(conditions) ⇒ Object

Same check for flag conditions, which may also reference audiences.



334
335
336
337
338
339
 
# File 'lib/lucerna/gates/evaluate.rb', line 334

def unsupported_flag_operator(conditions)
  conditions.find do |condition|
    !OPERATORS.include?(condition["operator"]) &&
      !AUDIENCE_OPERATORS.include?(condition["operator"])
  end&.fetch("operator")
end

.unsupported_operator(conditions) ⇒ Object

The first operator in the list this engine doesn't know, if any.



323
324
325
 
# File 'lib/lucerna/gates/evaluate.rb', line 323

def unsupported_operator(conditions)
  conditions.find { |condition| !OPERATORS.include?(condition["operator"]) }&.fetch("operator")
end